blob: fe3119a71adacb65172cdc6be3c390f55e234679 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/ext3/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
23#include <linux/jbd.h>
24#include <linux/ext3_fs.h>
25#include <linux/ext3_jbd.h>
26#include <linux/slab.h>
27#include <linux/init.h>
28#include <linux/blkdev.h>
29#include <linux/parser.h>
30#include <linux/smp_lock.h>
31#include <linux/buffer_head.h>
Christoph Hellwiga5694252007-07-17 04:04:28 -070032#include <linux/exportfs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033#include <linux/vfs.h>
34#include <linux/random.h>
35#include <linux/mount.h>
36#include <linux/namei.h>
37#include <linux/quotaops.h>
Mark Bellon8fc27512005-09-06 15:16:54 -070038#include <linux/seq_file.h>
vignesh babu3fc74262007-07-15 23:41:17 -070039#include <linux/log2.h>
Ben Dooks381be2542005-10-30 15:02:56 -080040
Linus Torvalds1da177e2005-04-16 15:20:36 -070041#include <asm/uaccess.h>
Ben Dooks381be2542005-10-30 15:02:56 -080042
Linus Torvalds1da177e2005-04-16 15:20:36 -070043#include "xattr.h"
44#include "acl.h"
Ben Dooks381be2542005-10-30 15:02:56 -080045#include "namei.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070046
Johann Lombardi71b96252006-01-08 01:03:20 -080047static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
48 unsigned long journal_devnum);
Linus Torvalds1da177e2005-04-16 15:20:36 -070049static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
Eric Sandeeneee194e2006-09-27 01:49:30 -070050 unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -070051static void ext3_commit_super (struct super_block * sb,
52 struct ext3_super_block * es,
53 int sync);
54static void ext3_mark_recovery_complete(struct super_block * sb,
55 struct ext3_super_block * es);
56static void ext3_clear_journal_err(struct super_block * sb,
57 struct ext3_super_block * es);
58static int ext3_sync_fs(struct super_block *sb, int wait);
59static const char *ext3_decode_error(struct super_block * sb, int errno,
60 char nbuf[16]);
61static int ext3_remount (struct super_block * sb, int * flags, char * data);
David Howells726c3342006-06-23 02:02:58 -070062static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
Linus Torvalds1da177e2005-04-16 15:20:36 -070063static void ext3_unlockfs(struct super_block *sb);
64static void ext3_write_super (struct super_block * sb);
65static void ext3_write_super_lockfs(struct super_block *sb);
66
Mingming Caoae6ddcc2006-09-27 01:49:27 -070067/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070068 * Wrappers for journal_start/end.
69 *
70 * The only special thing we need to do here is to make sure that all
71 * journal_end calls result in the superblock being marked dirty, so
72 * that sync() will call the filesystem's write_super callback if
Mingming Caoae6ddcc2006-09-27 01:49:27 -070073 * appropriate.
Linus Torvalds1da177e2005-04-16 15:20:36 -070074 */
75handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
76{
77 journal_t *journal;
78
79 if (sb->s_flags & MS_RDONLY)
80 return ERR_PTR(-EROFS);
81
82 /* Special case here: if the journal has aborted behind our
83 * backs (eg. EIO in the commit thread), then we still need to
84 * take the FS itself readonly cleanly. */
85 journal = EXT3_SB(sb)->s_journal;
86 if (is_journal_aborted(journal)) {
Harvey Harrisone05b6b52008-04-28 02:16:15 -070087 ext3_abort(sb, __func__,
Linus Torvalds1da177e2005-04-16 15:20:36 -070088 "Detected aborted journal");
89 return ERR_PTR(-EROFS);
90 }
91
92 return journal_start(journal, nblocks);
93}
94
Mingming Caoae6ddcc2006-09-27 01:49:27 -070095/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070096 * The only special thing we need to do here is to make sure that all
97 * journal_stop calls result in the superblock being marked dirty, so
98 * that sync() will call the filesystem's write_super callback if
Mingming Caoae6ddcc2006-09-27 01:49:27 -070099 * appropriate.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100 */
101int __ext3_journal_stop(const char *where, handle_t *handle)
102{
103 struct super_block *sb;
104 int err;
105 int rc;
106
107 sb = handle->h_transaction->t_journal->j_private;
108 err = handle->h_err;
109 rc = journal_stop(handle);
110
111 if (!err)
112 err = rc;
113 if (err)
114 __ext3_std_error(sb, where, err);
115 return err;
116}
117
118void ext3_journal_abort_handle(const char *caller, const char *err_fn,
119 struct buffer_head *bh, handle_t *handle, int err)
120{
121 char nbuf[16];
122 const char *errstr = ext3_decode_error(NULL, err, nbuf);
123
124 if (bh)
125 BUFFER_TRACE(bh, "abort");
126
127 if (!handle->h_err)
128 handle->h_err = err;
129
130 if (is_handle_aborted(handle))
131 return;
132
133 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
134 caller, errstr, err_fn);
135
136 journal_abort_handle(handle);
137}
138
139/* Deal with the reporting of failure conditions on a filesystem such as
140 * inconsistencies detected or read IO failures.
141 *
142 * On ext2, we can store the error state of the filesystem in the
143 * superblock. That is not possible on ext3, because we may have other
144 * write ordering constraints on the superblock which prevent us from
145 * writing it out straight away; and given that the journal is about to
146 * be aborted, we can't rely on the current, or future, transactions to
147 * write out the superblock safely.
148 *
149 * We'll just use the journal_abort() error code to record an error in
150 * the journal instead. On recovery, the journal will compain about
151 * that error until we've noted it down and cleared it.
152 */
153
154static void ext3_handle_error(struct super_block *sb)
155{
156 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
157
158 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
159 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
160
161 if (sb->s_flags & MS_RDONLY)
162 return;
163
Vasily Averin7543fc72006-09-27 01:49:33 -0700164 if (!test_opt (sb, ERRORS_CONT)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165 journal_t *journal = EXT3_SB(sb)->s_journal;
166
167 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
168 if (journal)
169 journal_abort(journal, -EIO);
170 }
Vasily Averin7543fc72006-09-27 01:49:33 -0700171 if (test_opt (sb, ERRORS_RO)) {
172 printk (KERN_CRIT "Remounting filesystem read-only\n");
173 sb->s_flags |= MS_RDONLY;
174 }
175 ext3_commit_super(sb, es, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176 if (test_opt(sb, ERRORS_PANIC))
177 panic("EXT3-fs (device %s): panic forced after error\n",
178 sb->s_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179}
180
181void ext3_error (struct super_block * sb, const char * function,
182 const char * fmt, ...)
183{
184 va_list args;
185
186 va_start(args, fmt);
187 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
188 vprintk(fmt, args);
189 printk("\n");
190 va_end(args);
191
192 ext3_handle_error(sb);
193}
194
195static const char *ext3_decode_error(struct super_block * sb, int errno,
196 char nbuf[16])
197{
198 char *errstr = NULL;
199
200 switch (errno) {
201 case -EIO:
202 errstr = "IO failure";
203 break;
204 case -ENOMEM:
205 errstr = "Out of memory";
206 break;
207 case -EROFS:
208 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
209 errstr = "Journal has aborted";
210 else
211 errstr = "Readonly filesystem";
212 break;
213 default:
214 /* If the caller passed in an extra buffer for unknown
215 * errors, textualise them now. Else we just return
216 * NULL. */
217 if (nbuf) {
218 /* Check for truncated error codes... */
219 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
220 errstr = nbuf;
221 }
222 break;
223 }
224
225 return errstr;
226}
227
228/* __ext3_std_error decodes expected errors from journaling functions
229 * automatically and invokes the appropriate error response. */
230
231void __ext3_std_error (struct super_block * sb, const char * function,
232 int errno)
233{
234 char nbuf[16];
Stephen Tweedie30121622005-05-18 11:47:17 -0400235 const char *errstr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236
Stephen Tweedie30121622005-05-18 11:47:17 -0400237 /* Special case: if the error is EROFS, and we're not already
238 * inside a transaction, then there's really no point in logging
239 * an error. */
240 if (errno == -EROFS && journal_current_handle() == NULL &&
241 (sb->s_flags & MS_RDONLY))
242 return;
243
244 errstr = ext3_decode_error(sb, errno, nbuf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245 printk (KERN_CRIT "EXT3-fs error (device %s) in %s: %s\n",
246 sb->s_id, function, errstr);
247
248 ext3_handle_error(sb);
249}
250
251/*
252 * ext3_abort is a much stronger failure handler than ext3_error. The
253 * abort function may be used to deal with unrecoverable failures such
254 * as journal IO errors or ENOMEM at a critical moment in log management.
255 *
256 * We unconditionally force the filesystem into an ABORT|READONLY state,
257 * unless the error response on the fs has been set to panic in which
258 * case we take the easy way out and panic immediately.
259 */
260
261void ext3_abort (struct super_block * sb, const char * function,
262 const char * fmt, ...)
263{
264 va_list args;
265
266 printk (KERN_CRIT "ext3_abort called.\n");
267
268 va_start(args, fmt);
269 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
270 vprintk(fmt, args);
271 printk("\n");
272 va_end(args);
273
274 if (test_opt(sb, ERRORS_PANIC))
275 panic("EXT3-fs panic from previous error\n");
276
277 if (sb->s_flags & MS_RDONLY)
278 return;
279
280 printk(KERN_CRIT "Remounting filesystem read-only\n");
281 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
282 sb->s_flags |= MS_RDONLY;
283 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
284 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
285}
286
287void ext3_warning (struct super_block * sb, const char * function,
288 const char * fmt, ...)
289{
290 va_list args;
291
292 va_start(args, fmt);
293 printk(KERN_WARNING "EXT3-fs warning (device %s): %s: ",
294 sb->s_id, function);
295 vprintk(fmt, args);
296 printk("\n");
297 va_end(args);
298}
299
300void ext3_update_dynamic_rev(struct super_block *sb)
301{
302 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
303
304 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
305 return;
306
Harvey Harrisone05b6b52008-04-28 02:16:15 -0700307 ext3_warning(sb, __func__,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308 "updating to rev %d because of new feature flag, "
309 "running e2fsck is recommended",
310 EXT3_DYNAMIC_REV);
311
312 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
313 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
314 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
315 /* leave es->s_feature_*compat flags alone */
316 /* es->s_uuid will be set by e2fsck if empty */
317
318 /*
319 * The rest of the superblock fields should be zero, and if not it
320 * means they are likely already in use, so leave them alone. We
321 * can leave it up to e2fsck to clean up any inconsistencies there.
322 */
323}
324
325/*
326 * Open the external journal device
327 */
328static struct block_device *ext3_blkdev_get(dev_t dev)
329{
330 struct block_device *bdev;
331 char b[BDEVNAME_SIZE];
332
333 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
334 if (IS_ERR(bdev))
335 goto fail;
336 return bdev;
337
338fail:
339 printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
340 __bdevname(dev, b), PTR_ERR(bdev));
341 return NULL;
342}
343
344/*
345 * Release the journal device
346 */
347static int ext3_blkdev_put(struct block_device *bdev)
348{
349 bd_release(bdev);
350 return blkdev_put(bdev);
351}
352
353static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
354{
355 struct block_device *bdev;
356 int ret = -ENODEV;
357
358 bdev = sbi->journal_bdev;
359 if (bdev) {
360 ret = ext3_blkdev_put(bdev);
361 sbi->journal_bdev = NULL;
362 }
363 return ret;
364}
365
366static inline struct inode *orphan_list_entry(struct list_head *l)
367{
368 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
369}
370
371static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
372{
373 struct list_head *l;
374
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700375 printk(KERN_ERR "sb orphan head is %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376 le32_to_cpu(sbi->s_es->s_last_orphan));
377
378 printk(KERN_ERR "sb_info orphan list:\n");
379 list_for_each(l, &sbi->s_orphan) {
380 struct inode *inode = orphan_list_entry(l);
381 printk(KERN_ERR " "
Eric Sandeeneee194e2006-09-27 01:49:30 -0700382 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 inode->i_sb->s_id, inode->i_ino, inode,
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700384 inode->i_mode, inode->i_nlink,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 NEXT_ORPHAN(inode));
386 }
387}
388
389static void ext3_put_super (struct super_block * sb)
390{
391 struct ext3_sb_info *sbi = EXT3_SB(sb);
392 struct ext3_super_block *es = sbi->s_es;
393 int i;
394
395 ext3_xattr_put_super(sb);
396 journal_destroy(sbi->s_journal);
397 if (!(sb->s_flags & MS_RDONLY)) {
398 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
399 es->s_state = cpu_to_le16(sbi->s_mount_state);
400 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
401 mark_buffer_dirty(sbi->s_sbh);
402 ext3_commit_super(sb, es, 1);
403 }
404
405 for (i = 0; i < sbi->s_gdb_count; i++)
406 brelse(sbi->s_group_desc[i]);
407 kfree(sbi->s_group_desc);
408 percpu_counter_destroy(&sbi->s_freeblocks_counter);
409 percpu_counter_destroy(&sbi->s_freeinodes_counter);
410 percpu_counter_destroy(&sbi->s_dirs_counter);
411 brelse(sbi->s_sbh);
412#ifdef CONFIG_QUOTA
413 for (i = 0; i < MAXQUOTAS; i++)
414 kfree(sbi->s_qf_names[i]);
415#endif
416
417 /* Debugging code just in case the in-memory inode orphan list
418 * isn't empty. The on-disk one can be non-empty if we've
419 * detected an error and taken the fs readonly, but the
420 * in-memory list had better be clean by this point. */
421 if (!list_empty(&sbi->s_orphan))
422 dump_orphan_list(sb, sbi);
423 J_ASSERT(list_empty(&sbi->s_orphan));
424
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700425 invalidate_bdev(sb->s_bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
427 /*
428 * Invalidate the journal device's buffers. We don't want them
429 * floating about in memory - the physical journal device may
430 * hotswapped, and it breaks the `ro-after' testing code.
431 */
432 sync_blockdev(sbi->journal_bdev);
Peter Zijlstraf98393a2007-05-06 14:49:54 -0700433 invalidate_bdev(sbi->journal_bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 ext3_blkdev_remove(sbi);
435 }
436 sb->s_fs_info = NULL;
437 kfree(sbi);
438 return;
439}
440
Christoph Lametere18b8902006-12-06 20:33:20 -0800441static struct kmem_cache *ext3_inode_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442
443/*
444 * Called inside transaction, so use GFP_NOFS
445 */
446static struct inode *ext3_alloc_inode(struct super_block *sb)
447{
448 struct ext3_inode_info *ei;
449
Christoph Lametere6b4f8d2006-12-06 20:33:14 -0800450 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451 if (!ei)
452 return NULL;
453#ifdef CONFIG_EXT3_FS_POSIX_ACL
454 ei->i_acl = EXT3_ACL_NOT_CACHED;
455 ei->i_default_acl = EXT3_ACL_NOT_CACHED;
456#endif
457 ei->i_block_alloc_info = NULL;
458 ei->vfs_inode.i_version = 1;
459 return &ei->vfs_inode;
460}
461
462static void ext3_destroy_inode(struct inode *inode)
463{
Vasily Averin9f7dd932007-07-15 23:40:45 -0700464 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
465 printk("EXT3 Inode %p: orphan list check failed!\n",
466 EXT3_I(inode));
467 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
468 EXT3_I(inode), sizeof(struct ext3_inode_info),
469 false);
470 dump_stack();
471 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
473}
474
Christoph Lameter4ba9b9d2007-10-16 23:25:51 -0700475static void init_once(struct kmem_cache * cachep, void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476{
477 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
478
Christoph Lametera35afb82007-05-16 22:10:57 -0700479 INIT_LIST_HEAD(&ei->i_orphan);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480#ifdef CONFIG_EXT3_FS_XATTR
Christoph Lametera35afb82007-05-16 22:10:57 -0700481 init_rwsem(&ei->xattr_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482#endif
Christoph Lametera35afb82007-05-16 22:10:57 -0700483 mutex_init(&ei->truncate_mutex);
484 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485}
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700486
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487static int init_inodecache(void)
488{
489 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
490 sizeof(struct ext3_inode_info),
Paul Jacksonfffb60f2006-03-24 03:16:06 -0800491 0, (SLAB_RECLAIM_ACCOUNT|
492 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900493 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 if (ext3_inode_cachep == NULL)
495 return -ENOMEM;
496 return 0;
497}
498
499static void destroy_inodecache(void)
500{
Alexey Dobriyan1a1d92c2006-09-27 01:49:40 -0700501 kmem_cache_destroy(ext3_inode_cachep);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502}
503
504static void ext3_clear_inode(struct inode *inode)
505{
506 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
507#ifdef CONFIG_EXT3_FS_POSIX_ACL
Andrew Mortone6022602006-06-23 02:05:32 -0700508 if (EXT3_I(inode)->i_acl &&
509 EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
510 posix_acl_release(EXT3_I(inode)->i_acl);
511 EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
512 }
513 if (EXT3_I(inode)->i_default_acl &&
514 EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
515 posix_acl_release(EXT3_I(inode)->i_default_acl);
516 EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
517 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518#endif
519 ext3_discard_reservation(inode);
520 EXT3_I(inode)->i_block_alloc_info = NULL;
Andrew Mortone6022602006-06-23 02:05:32 -0700521 if (unlikely(rsv))
522 kfree(rsv);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523}
524
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200525static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
Mark Bellon8fc27512005-09-06 15:16:54 -0700526{
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200527#if defined(CONFIG_QUOTA)
OGAWA Hirofumi275abf52005-09-22 21:44:03 -0700528 struct ext3_sb_info *sbi = EXT3_SB(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529
Mark Bellon8fc27512005-09-06 15:16:54 -0700530 if (sbi->s_jquota_fmt)
531 seq_printf(seq, ",jqfmt=%s",
532 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
533
534 if (sbi->s_qf_names[USRQUOTA])
535 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
536
537 if (sbi->s_qf_names[GRPQUOTA])
538 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
539
540 if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
541 seq_puts(seq, ",usrquota");
542
543 if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
544 seq_puts(seq, ",grpquota");
545#endif
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200546}
547
Miklos Szeredi571beed2007-10-16 23:26:26 -0700548/*
549 * Show an option if
550 * - it's set to a non-default value OR
551 * - if the per-sb default is different from the global default
552 */
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200553static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
554{
555 struct super_block *sb = vfs->mnt_sb;
Miklos Szeredi571beed2007-10-16 23:26:26 -0700556 struct ext3_sb_info *sbi = EXT3_SB(sb);
557 struct ext3_super_block *es = sbi->s_es;
558 unsigned long def_mount_opts;
559
560 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
561
562 if (sbi->s_sb_block != 1)
563 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
564 if (test_opt(sb, MINIX_DF))
565 seq_puts(seq, ",minixdf");
566 if (test_opt(sb, GRPID))
567 seq_puts(seq, ",grpid");
568 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
569 seq_puts(seq, ",nogrpid");
570 if (sbi->s_resuid != EXT3_DEF_RESUID ||
571 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
572 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
573 }
574 if (sbi->s_resgid != EXT3_DEF_RESGID ||
575 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
576 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
577 }
Aneesh Kumar K.V1eca93f2008-02-06 01:40:14 -0800578 if (test_opt(sb, ERRORS_RO)) {
Miklos Szeredi571beed2007-10-16 23:26:26 -0700579 int def_errors = le16_to_cpu(es->s_errors);
580
581 if (def_errors == EXT3_ERRORS_PANIC ||
Aneesh Kumar K.V1eca93f2008-02-06 01:40:14 -0800582 def_errors == EXT3_ERRORS_CONTINUE) {
583 seq_puts(seq, ",errors=remount-ro");
Miklos Szeredi571beed2007-10-16 23:26:26 -0700584 }
585 }
Aneesh Kumar K.V1eca93f2008-02-06 01:40:14 -0800586 if (test_opt(sb, ERRORS_CONT))
587 seq_puts(seq, ",errors=continue");
Miklos Szeredi571beed2007-10-16 23:26:26 -0700588 if (test_opt(sb, ERRORS_PANIC))
589 seq_puts(seq, ",errors=panic");
590 if (test_opt(sb, NO_UID32))
591 seq_puts(seq, ",nouid32");
592 if (test_opt(sb, DEBUG))
593 seq_puts(seq, ",debug");
594 if (test_opt(sb, OLDALLOC))
595 seq_puts(seq, ",oldalloc");
596#ifdef CONFIG_EXT3_FS_XATTR
597 if (test_opt(sb, XATTR_USER))
598 seq_puts(seq, ",user_xattr");
599 if (!test_opt(sb, XATTR_USER) &&
600 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
601 seq_puts(seq, ",nouser_xattr");
602 }
603#endif
604#ifdef CONFIG_EXT3_FS_POSIX_ACL
605 if (test_opt(sb, POSIX_ACL))
606 seq_puts(seq, ",acl");
607 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
608 seq_puts(seq, ",noacl");
609#endif
610 if (!test_opt(sb, RESERVATION))
611 seq_puts(seq, ",noreservation");
612 if (sbi->s_commit_interval) {
613 seq_printf(seq, ",commit=%u",
614 (unsigned) (sbi->s_commit_interval / HZ));
615 }
616 if (test_opt(sb, BARRIER))
617 seq_puts(seq, ",barrier=1");
618 if (test_opt(sb, NOBH))
619 seq_puts(seq, ",nobh");
Peter Osterlund6cd37cd2005-10-28 20:23:39 +0200620
621 if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
622 seq_puts(seq, ",data=journal");
623 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
624 seq_puts(seq, ",data=ordered");
625 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
626 seq_puts(seq, ",data=writeback");
627
628 ext3_show_quota_options(seq, sb);
Mark Bellon8fc27512005-09-06 15:16:54 -0700629
630 return 0;
631}
632
NeilBrownfdb36672006-09-16 12:15:38 -0700633
Christoph Hellwig74af0ba2007-10-21 16:42:07 -0700634static struct inode *ext3_nfs_get_inode(struct super_block *sb,
635 u64 ino, u32 generation)
NeilBrownfdb36672006-09-16 12:15:38 -0700636{
NeilBrownfdb36672006-09-16 12:15:38 -0700637 struct inode *inode;
NeilBrownfdb36672006-09-16 12:15:38 -0700638
639 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
640 return ERR_PTR(-ESTALE);
641 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
642 return ERR_PTR(-ESTALE);
643
644 /* iget isn't really right if the inode is currently unallocated!!
645 *
646 * ext3_read_inode will return a bad_inode if the inode had been
647 * deleted, so we should be safe.
648 *
649 * Currently we don't know the generation for parent directory, so
650 * a generation of 0 means "accept any"
651 */
David Howells473043d2008-02-07 00:15:36 -0800652 inode = ext3_iget(sb, ino);
653 if (IS_ERR(inode))
654 return ERR_CAST(inode);
655 if (generation && inode->i_generation != generation) {
NeilBrownfdb36672006-09-16 12:15:38 -0700656 iput(inode);
657 return ERR_PTR(-ESTALE);
658 }
Christoph Hellwig74af0ba2007-10-21 16:42:07 -0700659
660 return inode;
661}
662
663static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
664 int fh_len, int fh_type)
665{
666 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
667 ext3_nfs_get_inode);
668}
669
670static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
671 int fh_len, int fh_type)
672{
673 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
674 ext3_nfs_get_inode);
NeilBrownfdb36672006-09-16 12:15:38 -0700675}
676
Mark Bellon8fc27512005-09-06 15:16:54 -0700677#ifdef CONFIG_QUOTA
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
679#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
680
681static int ext3_dquot_initialize(struct inode *inode, int type);
682static int ext3_dquot_drop(struct inode *inode);
683static int ext3_write_dquot(struct dquot *dquot);
684static int ext3_acquire_dquot(struct dquot *dquot);
685static int ext3_release_dquot(struct dquot *dquot);
686static int ext3_mark_dquot_dirty(struct dquot *dquot);
687static int ext3_write_info(struct super_block *sb, int type);
Jan Kara2fd83a42008-04-28 02:14:34 -0700688static int ext3_quota_on(struct super_block *sb, int type, int format_id,
689 char *path, int remount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690static int ext3_quota_on_mount(struct super_block *sb, int type);
691static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
692 size_t len, loff_t off);
693static ssize_t ext3_quota_write(struct super_block *sb, int type,
694 const char *data, size_t len, loff_t off);
695
696static struct dquot_operations ext3_quota_operations = {
697 .initialize = ext3_dquot_initialize,
698 .drop = ext3_dquot_drop,
699 .alloc_space = dquot_alloc_space,
700 .alloc_inode = dquot_alloc_inode,
701 .free_space = dquot_free_space,
702 .free_inode = dquot_free_inode,
703 .transfer = dquot_transfer,
704 .write_dquot = ext3_write_dquot,
705 .acquire_dquot = ext3_acquire_dquot,
706 .release_dquot = ext3_release_dquot,
707 .mark_dirty = ext3_mark_dquot_dirty,
708 .write_info = ext3_write_info
709};
710
711static struct quotactl_ops ext3_qctl_operations = {
712 .quota_on = ext3_quota_on,
713 .quota_off = vfs_quota_off,
714 .quota_sync = vfs_quota_sync,
715 .get_info = vfs_get_dqinfo,
716 .set_info = vfs_set_dqinfo,
717 .get_dqblk = vfs_get_dqblk,
718 .set_dqblk = vfs_set_dqblk
719};
720#endif
721
Josef 'Jeff' Sipekee9b6d62007-02-12 00:55:41 -0800722static const struct super_operations ext3_sops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723 .alloc_inode = ext3_alloc_inode,
724 .destroy_inode = ext3_destroy_inode,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725 .write_inode = ext3_write_inode,
726 .dirty_inode = ext3_dirty_inode,
727 .delete_inode = ext3_delete_inode,
728 .put_super = ext3_put_super,
729 .write_super = ext3_write_super,
730 .sync_fs = ext3_sync_fs,
731 .write_super_lockfs = ext3_write_super_lockfs,
732 .unlockfs = ext3_unlockfs,
733 .statfs = ext3_statfs,
734 .remount_fs = ext3_remount,
735 .clear_inode = ext3_clear_inode,
Mark Bellon8fc27512005-09-06 15:16:54 -0700736 .show_options = ext3_show_options,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737#ifdef CONFIG_QUOTA
738 .quota_read = ext3_quota_read,
739 .quota_write = ext3_quota_write,
740#endif
741};
742
Christoph Hellwig39655162007-10-21 16:42:17 -0700743static const struct export_operations ext3_export_ops = {
Christoph Hellwig74af0ba2007-10-21 16:42:07 -0700744 .fh_to_dentry = ext3_fh_to_dentry,
745 .fh_to_parent = ext3_fh_to_parent,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746 .get_parent = ext3_get_parent,
747};
748
749enum {
750 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
751 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
Adrian Bunk2860b732005-11-08 21:34:59 -0800752 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
Badari Pulavartyade1a292006-06-26 00:25:04 -0700754 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
Johann Lombardi71b96252006-01-08 01:03:20 -0800755 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
757 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
Jan Kara1f545872005-06-23 22:01:04 -0700758 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
Mark Bellon8fc27512005-09-06 15:16:54 -0700759 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
760 Opt_grpquota
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761};
762
763static match_table_t tokens = {
764 {Opt_bsd_df, "bsddf"},
765 {Opt_minix_df, "minixdf"},
766 {Opt_grpid, "grpid"},
767 {Opt_grpid, "bsdgroups"},
768 {Opt_nogrpid, "nogrpid"},
769 {Opt_nogrpid, "sysvgroups"},
770 {Opt_resgid, "resgid=%u"},
771 {Opt_resuid, "resuid=%u"},
772 {Opt_sb, "sb=%u"},
773 {Opt_err_cont, "errors=continue"},
774 {Opt_err_panic, "errors=panic"},
775 {Opt_err_ro, "errors=remount-ro"},
776 {Opt_nouid32, "nouid32"},
777 {Opt_nocheck, "nocheck"},
778 {Opt_nocheck, "check=none"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 {Opt_debug, "debug"},
780 {Opt_oldalloc, "oldalloc"},
781 {Opt_orlov, "orlov"},
782 {Opt_user_xattr, "user_xattr"},
783 {Opt_nouser_xattr, "nouser_xattr"},
784 {Opt_acl, "acl"},
785 {Opt_noacl, "noacl"},
786 {Opt_reservation, "reservation"},
787 {Opt_noreservation, "noreservation"},
788 {Opt_noload, "noload"},
789 {Opt_nobh, "nobh"},
Badari Pulavartyade1a292006-06-26 00:25:04 -0700790 {Opt_bh, "bh"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 {Opt_commit, "commit=%u"},
792 {Opt_journal_update, "journal=update"},
793 {Opt_journal_inum, "journal=%u"},
Johann Lombardi71b96252006-01-08 01:03:20 -0800794 {Opt_journal_dev, "journal_dev=%u"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795 {Opt_abort, "abort"},
796 {Opt_data_journal, "data=journal"},
797 {Opt_data_ordered, "data=ordered"},
798 {Opt_data_writeback, "data=writeback"},
799 {Opt_offusrjquota, "usrjquota="},
800 {Opt_usrjquota, "usrjquota=%s"},
801 {Opt_offgrpjquota, "grpjquota="},
802 {Opt_grpjquota, "grpjquota=%s"},
803 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
804 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
Mark Bellon8fc27512005-09-06 15:16:54 -0700805 {Opt_grpquota, "grpquota"},
Jan Kara1f545872005-06-23 22:01:04 -0700806 {Opt_noquota, "noquota"},
807 {Opt_quota, "quota"},
Mark Bellon8fc27512005-09-06 15:16:54 -0700808 {Opt_usrquota, "usrquota"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809 {Opt_barrier, "barrier=%u"},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 {Opt_resize, "resize"},
Josef Bacik925872162008-03-04 14:29:43 -0800811 {Opt_err, NULL},
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812};
813
Mingming Cao43d23f92006-06-25 05:48:07 -0700814static ext3_fsblk_t get_sb_block(void **data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815{
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700816 ext3_fsblk_t sb_block;
817 char *options = (char *) *data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818
819 if (!options || strncmp(options, "sb=", 3) != 0)
820 return 1; /* Default location */
821 options += 3;
Mingming Cao43d23f92006-06-25 05:48:07 -0700822 /*todo: use simple_strtoll with >32bit ext3 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823 sb_block = simple_strtoul(options, &options, 0);
824 if (*options && *options != ',') {
825 printk("EXT3-fs: Invalid sb specification: %s\n",
826 (char *) *data);
827 return 1;
828 }
829 if (*options == ',')
830 options++;
831 *data = (void *) options;
832 return sb_block;
833}
834
Johann Lombardi71b96252006-01-08 01:03:20 -0800835static int parse_options (char *options, struct super_block *sb,
Eric Sandeeneee194e2006-09-27 01:49:30 -0700836 unsigned int *inum, unsigned long *journal_devnum,
Mingming Cao43d23f92006-06-25 05:48:07 -0700837 ext3_fsblk_t *n_blocks_count, int is_remount)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838{
839 struct ext3_sb_info *sbi = EXT3_SB(sb);
840 char * p;
841 substring_t args[MAX_OPT_ARGS];
842 int data_opt = 0;
843 int option;
844#ifdef CONFIG_QUOTA
845 int qtype;
846 char *qname;
847#endif
848
849 if (!options)
850 return 1;
851
852 while ((p = strsep (&options, ",")) != NULL) {
853 int token;
854 if (!*p)
855 continue;
856
857 token = match_token(p, tokens, args);
858 switch (token) {
859 case Opt_bsd_df:
860 clear_opt (sbi->s_mount_opt, MINIX_DF);
861 break;
862 case Opt_minix_df:
863 set_opt (sbi->s_mount_opt, MINIX_DF);
864 break;
865 case Opt_grpid:
866 set_opt (sbi->s_mount_opt, GRPID);
867 break;
868 case Opt_nogrpid:
869 clear_opt (sbi->s_mount_opt, GRPID);
870 break;
871 case Opt_resuid:
872 if (match_int(&args[0], &option))
873 return 0;
874 sbi->s_resuid = option;
875 break;
876 case Opt_resgid:
877 if (match_int(&args[0], &option))
878 return 0;
879 sbi->s_resgid = option;
880 break;
881 case Opt_sb:
882 /* handled by get_sb_block() instead of here */
883 /* *sb_block = match_int(&args[0]); */
884 break;
885 case Opt_err_panic:
886 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
887 clear_opt (sbi->s_mount_opt, ERRORS_RO);
888 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
889 break;
890 case Opt_err_ro:
891 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
892 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
893 set_opt (sbi->s_mount_opt, ERRORS_RO);
894 break;
895 case Opt_err_cont:
896 clear_opt (sbi->s_mount_opt, ERRORS_RO);
897 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
898 set_opt (sbi->s_mount_opt, ERRORS_CONT);
899 break;
900 case Opt_nouid32:
901 set_opt (sbi->s_mount_opt, NO_UID32);
902 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 case Opt_nocheck:
904 clear_opt (sbi->s_mount_opt, CHECK);
905 break;
906 case Opt_debug:
907 set_opt (sbi->s_mount_opt, DEBUG);
908 break;
909 case Opt_oldalloc:
910 set_opt (sbi->s_mount_opt, OLDALLOC);
911 break;
912 case Opt_orlov:
913 clear_opt (sbi->s_mount_opt, OLDALLOC);
914 break;
915#ifdef CONFIG_EXT3_FS_XATTR
916 case Opt_user_xattr:
917 set_opt (sbi->s_mount_opt, XATTR_USER);
918 break;
919 case Opt_nouser_xattr:
920 clear_opt (sbi->s_mount_opt, XATTR_USER);
921 break;
922#else
923 case Opt_user_xattr:
924 case Opt_nouser_xattr:
925 printk("EXT3 (no)user_xattr options not supported\n");
926 break;
927#endif
928#ifdef CONFIG_EXT3_FS_POSIX_ACL
929 case Opt_acl:
930 set_opt(sbi->s_mount_opt, POSIX_ACL);
931 break;
932 case Opt_noacl:
933 clear_opt(sbi->s_mount_opt, POSIX_ACL);
934 break;
935#else
936 case Opt_acl:
937 case Opt_noacl:
938 printk("EXT3 (no)acl options not supported\n");
939 break;
940#endif
941 case Opt_reservation:
942 set_opt(sbi->s_mount_opt, RESERVATION);
943 break;
944 case Opt_noreservation:
945 clear_opt(sbi->s_mount_opt, RESERVATION);
946 break;
947 case Opt_journal_update:
948 /* @@@ FIXME */
949 /* Eventually we will want to be able to create
950 a journal file here. For now, only allow the
951 user to specify an existing inode to be the
952 journal file. */
953 if (is_remount) {
954 printk(KERN_ERR "EXT3-fs: cannot specify "
955 "journal on remount\n");
956 return 0;
957 }
958 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
959 break;
960 case Opt_journal_inum:
961 if (is_remount) {
962 printk(KERN_ERR "EXT3-fs: cannot specify "
963 "journal on remount\n");
964 return 0;
965 }
966 if (match_int(&args[0], &option))
967 return 0;
968 *inum = option;
969 break;
Johann Lombardi71b96252006-01-08 01:03:20 -0800970 case Opt_journal_dev:
971 if (is_remount) {
972 printk(KERN_ERR "EXT3-fs: cannot specify "
973 "journal on remount\n");
974 return 0;
975 }
976 if (match_int(&args[0], &option))
977 return 0;
978 *journal_devnum = option;
979 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980 case Opt_noload:
981 set_opt (sbi->s_mount_opt, NOLOAD);
982 break;
983 case Opt_commit:
984 if (match_int(&args[0], &option))
985 return 0;
986 if (option < 0)
987 return 0;
988 if (option == 0)
989 option = JBD_DEFAULT_MAX_COMMIT_AGE;
990 sbi->s_commit_interval = HZ * option;
991 break;
992 case Opt_data_journal:
993 data_opt = EXT3_MOUNT_JOURNAL_DATA;
994 goto datacheck;
995 case Opt_data_ordered:
996 data_opt = EXT3_MOUNT_ORDERED_DATA;
997 goto datacheck;
998 case Opt_data_writeback:
999 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1000 datacheck:
1001 if (is_remount) {
1002 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
1003 != data_opt) {
1004 printk(KERN_ERR
1005 "EXT3-fs: cannot change data "
1006 "mode on remount\n");
1007 return 0;
1008 }
1009 } else {
1010 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
1011 sbi->s_mount_opt |= data_opt;
1012 }
1013 break;
1014#ifdef CONFIG_QUOTA
1015 case Opt_usrjquota:
1016 qtype = USRQUOTA;
1017 goto set_qf_name;
1018 case Opt_grpjquota:
1019 qtype = GRPQUOTA;
1020set_qf_name:
1021 if (sb_any_quota_enabled(sb)) {
1022 printk(KERN_ERR
1023 "EXT3-fs: Cannot change journalled "
1024 "quota options when quota turned on.\n");
1025 return 0;
1026 }
1027 qname = match_strdup(&args[0]);
1028 if (!qname) {
1029 printk(KERN_ERR
1030 "EXT3-fs: not enough memory for "
1031 "storing quotafile name.\n");
1032 return 0;
1033 }
1034 if (sbi->s_qf_names[qtype] &&
1035 strcmp(sbi->s_qf_names[qtype], qname)) {
1036 printk(KERN_ERR
1037 "EXT3-fs: %s quota file already "
1038 "specified.\n", QTYPE2NAME(qtype));
1039 kfree(qname);
1040 return 0;
1041 }
1042 sbi->s_qf_names[qtype] = qname;
1043 if (strchr(sbi->s_qf_names[qtype], '/')) {
1044 printk(KERN_ERR
1045 "EXT3-fs: quotafile must be on "
1046 "filesystem root.\n");
1047 kfree(sbi->s_qf_names[qtype]);
1048 sbi->s_qf_names[qtype] = NULL;
1049 return 0;
1050 }
Jan Kara1f545872005-06-23 22:01:04 -07001051 set_opt(sbi->s_mount_opt, QUOTA);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052 break;
1053 case Opt_offusrjquota:
1054 qtype = USRQUOTA;
1055 goto clear_qf_name;
1056 case Opt_offgrpjquota:
1057 qtype = GRPQUOTA;
1058clear_qf_name:
1059 if (sb_any_quota_enabled(sb)) {
1060 printk(KERN_ERR "EXT3-fs: Cannot change "
1061 "journalled quota options when "
1062 "quota turned on.\n");
1063 return 0;
1064 }
Jan Kara08c6a962005-07-12 13:58:28 -07001065 /*
1066 * The space will be released later when all options
1067 * are confirmed to be correct
1068 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 sbi->s_qf_names[qtype] = NULL;
1070 break;
1071 case Opt_jqfmt_vfsold:
1072 sbi->s_jquota_fmt = QFMT_VFS_OLD;
1073 break;
1074 case Opt_jqfmt_vfsv0:
1075 sbi->s_jquota_fmt = QFMT_VFS_V0;
1076 break;
Jan Kara1f545872005-06-23 22:01:04 -07001077 case Opt_quota:
Mark Bellon8fc27512005-09-06 15:16:54 -07001078 case Opt_usrquota:
Jan Kara1f545872005-06-23 22:01:04 -07001079 set_opt(sbi->s_mount_opt, QUOTA);
Mark Bellon8fc27512005-09-06 15:16:54 -07001080 set_opt(sbi->s_mount_opt, USRQUOTA);
1081 break;
1082 case Opt_grpquota:
1083 set_opt(sbi->s_mount_opt, QUOTA);
1084 set_opt(sbi->s_mount_opt, GRPQUOTA);
Jan Kara1f545872005-06-23 22:01:04 -07001085 break;
1086 case Opt_noquota:
1087 if (sb_any_quota_enabled(sb)) {
1088 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1089 "options when quota turned on.\n");
1090 return 0;
1091 }
1092 clear_opt(sbi->s_mount_opt, QUOTA);
Mark Bellon8fc27512005-09-06 15:16:54 -07001093 clear_opt(sbi->s_mount_opt, USRQUOTA);
1094 clear_opt(sbi->s_mount_opt, GRPQUOTA);
Jan Kara1f545872005-06-23 22:01:04 -07001095 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096#else
Mark Bellon8fc27512005-09-06 15:16:54 -07001097 case Opt_quota:
1098 case Opt_usrquota:
1099 case Opt_grpquota:
Jan Karafa1ff1e2008-04-28 02:16:14 -07001100 printk(KERN_ERR
1101 "EXT3-fs: quota options not supported.\n");
1102 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103 case Opt_usrjquota:
1104 case Opt_grpjquota:
1105 case Opt_offusrjquota:
1106 case Opt_offgrpjquota:
1107 case Opt_jqfmt_vfsold:
1108 case Opt_jqfmt_vfsv0:
1109 printk(KERN_ERR
Jan Karafa1ff1e2008-04-28 02:16:14 -07001110 "EXT3-fs: journaled quota options not "
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111 "supported.\n");
1112 break;
Jan Kara1f545872005-06-23 22:01:04 -07001113 case Opt_noquota:
1114 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115#endif
1116 case Opt_abort:
1117 set_opt(sbi->s_mount_opt, ABORT);
1118 break;
1119 case Opt_barrier:
1120 if (match_int(&args[0], &option))
1121 return 0;
1122 if (option)
1123 set_opt(sbi->s_mount_opt, BARRIER);
1124 else
1125 clear_opt(sbi->s_mount_opt, BARRIER);
1126 break;
1127 case Opt_ignore:
1128 break;
1129 case Opt_resize:
Jan Kara08c6a962005-07-12 13:58:28 -07001130 if (!is_remount) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131 printk("EXT3-fs: resize option only available "
1132 "for remount\n");
1133 return 0;
1134 }
KAMBAROV, ZAURc7f17212005-06-28 20:45:11 -07001135 if (match_int(&args[0], &option) != 0)
1136 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137 *n_blocks_count = option;
1138 break;
1139 case Opt_nobh:
1140 set_opt(sbi->s_mount_opt, NOBH);
1141 break;
Badari Pulavartyade1a292006-06-26 00:25:04 -07001142 case Opt_bh:
1143 clear_opt(sbi->s_mount_opt, NOBH);
1144 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145 default:
1146 printk (KERN_ERR
1147 "EXT3-fs: Unrecognized mount option \"%s\" "
1148 "or missing value\n", p);
1149 return 0;
1150 }
1151 }
1152#ifdef CONFIG_QUOTA
Mark Bellon8fc27512005-09-06 15:16:54 -07001153 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1154 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1155 sbi->s_qf_names[USRQUOTA])
1156 clear_opt(sbi->s_mount_opt, USRQUOTA);
1157
1158 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1159 sbi->s_qf_names[GRPQUOTA])
1160 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1161
1162 if ((sbi->s_qf_names[USRQUOTA] &&
1163 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1164 (sbi->s_qf_names[GRPQUOTA] &&
1165 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1166 printk(KERN_ERR "EXT3-fs: old and new quota "
1167 "format mixing.\n");
1168 return 0;
1169 }
1170
1171 if (!sbi->s_jquota_fmt) {
1172 printk(KERN_ERR "EXT3-fs: journalled quota format "
1173 "not specified.\n");
1174 return 0;
1175 }
1176 } else {
1177 if (sbi->s_jquota_fmt) {
1178 printk(KERN_ERR "EXT3-fs: journalled quota format "
1179 "specified with no journalling "
1180 "enabled.\n");
1181 return 0;
1182 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183 }
1184#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185 return 1;
1186}
1187
1188static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1189 int read_only)
1190{
1191 struct ext3_sb_info *sbi = EXT3_SB(sb);
1192 int res = 0;
1193
1194 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1195 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1196 "forcing read-only mode\n");
1197 res = MS_RDONLY;
1198 }
1199 if (read_only)
1200 return res;
1201 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1202 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1203 "running e2fsck is recommended\n");
1204 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1205 printk (KERN_WARNING
1206 "EXT3-fs warning: mounting fs with errors, "
1207 "running e2fsck is recommended\n");
1208 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1209 le16_to_cpu(es->s_mnt_count) >=
1210 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1211 printk (KERN_WARNING
1212 "EXT3-fs warning: maximal mount count reached, "
1213 "running e2fsck is recommended\n");
1214 else if (le32_to_cpu(es->s_checkinterval) &&
1215 (le32_to_cpu(es->s_lastcheck) +
1216 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1217 printk (KERN_WARNING
1218 "EXT3-fs warning: checktime reached, "
1219 "running e2fsck is recommended\n");
1220#if 0
1221 /* @@@ We _will_ want to clear the valid bit if we find
1222 inconsistencies, to force a fsck at reboot. But for
1223 a plain journaled filesystem we can keep it set as
1224 valid forever! :) */
Marcin Slusarze7f23eb2008-04-28 02:16:06 -07001225 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226#endif
1227 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1228 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
Marcin Slusarz50e8a282008-02-08 04:20:13 -08001229 le16_add_cpu(&es->s_mnt_count, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230 es->s_mtime = cpu_to_le32(get_seconds());
1231 ext3_update_dynamic_rev(sb);
1232 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1233
1234 ext3_commit_super(sb, es, 1);
1235 if (test_opt(sb, DEBUG))
1236 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1237 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1238 sb->s_blocksize,
1239 sbi->s_groups_count,
1240 EXT3_BLOCKS_PER_GROUP(sb),
1241 EXT3_INODES_PER_GROUP(sb),
1242 sbi->s_mount_opt);
1243
1244 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1245 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1246 char b[BDEVNAME_SIZE];
1247
1248 printk("external journal on %s\n",
1249 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1250 } else {
1251 printk("internal journal\n");
1252 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253 return res;
1254}
1255
1256/* Called at mount-time, super-block is locked */
Akinobu Mita197cd65a2008-02-06 01:40:16 -08001257static int ext3_check_descriptors(struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258{
1259 struct ext3_sb_info *sbi = EXT3_SB(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260 int i;
1261
1262 ext3_debug ("Checking group descriptors");
1263
Akinobu Mita197cd65a2008-02-06 01:40:16 -08001264 for (i = 0; i < sbi->s_groups_count; i++) {
1265 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
Akinobu Mita1eaafea2008-04-28 02:16:07 -07001266 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1267 ext3_fsblk_t last_block;
Akinobu Mita197cd65a2008-02-06 01:40:16 -08001268
Eric Sandeen855565e2006-09-27 01:49:29 -07001269 if (i == sbi->s_groups_count - 1)
1270 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1271 else
1272 last_block = first_block +
1273 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1274
Eric Sandeen855565e2006-09-27 01:49:29 -07001275 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1276 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277 {
1278 ext3_error (sb, "ext3_check_descriptors",
1279 "Block bitmap for group %d"
1280 " not in group (block %lu)!",
1281 i, (unsigned long)
1282 le32_to_cpu(gdp->bg_block_bitmap));
1283 return 0;
1284 }
Eric Sandeen855565e2006-09-27 01:49:29 -07001285 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1286 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287 {
1288 ext3_error (sb, "ext3_check_descriptors",
1289 "Inode bitmap for group %d"
1290 " not in group (block %lu)!",
1291 i, (unsigned long)
1292 le32_to_cpu(gdp->bg_inode_bitmap));
1293 return 0;
1294 }
Eric Sandeen855565e2006-09-27 01:49:29 -07001295 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
Eric Sandeen780dcdb2007-07-26 10:41:11 -07001296 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
Eric Sandeen855565e2006-09-27 01:49:29 -07001297 last_block)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298 {
1299 ext3_error (sb, "ext3_check_descriptors",
1300 "Inode table for group %d"
1301 " not in group (block %lu)!",
1302 i, (unsigned long)
1303 le32_to_cpu(gdp->bg_inode_table));
1304 return 0;
1305 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306 }
1307
1308 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1309 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1310 return 1;
1311}
1312
1313
1314/* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1315 * the superblock) which were deleted from all directories, but held open by
1316 * a process at the time of a crash. We walk the list and try to delete these
1317 * inodes at recovery time (only with a read-write filesystem).
1318 *
1319 * In order to keep the orphan inode chain consistent during traversal (in
1320 * case of crash during recovery), we link each inode into the superblock
1321 * orphan list_head and handle it the same way as an inode deletion during
1322 * normal operation (which journals the operations for us).
1323 *
1324 * We only do an iget() and an iput() on each inode, which is very safe if we
1325 * accidentally point at an in-use or already deleted inode. The worst that
1326 * can happen in this case is that we get a "bit already cleared" message from
1327 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1328 * e2fsck was run on this filesystem, and it must have already done the orphan
1329 * inode cleanup for us, so we can safely abort without any further action.
1330 */
1331static void ext3_orphan_cleanup (struct super_block * sb,
1332 struct ext3_super_block * es)
1333{
1334 unsigned int s_flags = sb->s_flags;
1335 int nr_orphans = 0, nr_truncates = 0;
1336#ifdef CONFIG_QUOTA
1337 int i;
1338#endif
1339 if (!es->s_last_orphan) {
1340 jbd_debug(4, "no orphan inodes to clean up\n");
1341 return;
1342 }
1343
Eric Sandeena8f48a92006-12-06 20:40:13 -08001344 if (bdev_read_only(sb->s_bdev)) {
1345 printk(KERN_ERR "EXT3-fs: write access "
1346 "unavailable, skipping orphan cleanup.\n");
1347 return;
1348 }
1349
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1351 if (es->s_last_orphan)
1352 jbd_debug(1, "Errors on filesystem, "
1353 "clearing orphan list.\n");
1354 es->s_last_orphan = 0;
1355 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1356 return;
1357 }
1358
1359 if (s_flags & MS_RDONLY) {
1360 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1361 sb->s_id);
1362 sb->s_flags &= ~MS_RDONLY;
1363 }
1364#ifdef CONFIG_QUOTA
1365 /* Needed for iput() to work correctly and not trash data */
1366 sb->s_flags |= MS_ACTIVE;
1367 /* Turn on quotas so that they are updated correctly */
1368 for (i = 0; i < MAXQUOTAS; i++) {
1369 if (EXT3_SB(sb)->s_qf_names[i]) {
1370 int ret = ext3_quota_on_mount(sb, i);
1371 if (ret < 0)
1372 printk(KERN_ERR
1373 "EXT3-fs: Cannot turn on journalled "
1374 "quota: error %d\n", ret);
1375 }
1376 }
1377#endif
1378
1379 while (es->s_last_orphan) {
1380 struct inode *inode;
1381
David Howells473043d2008-02-07 00:15:36 -08001382 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1383 if (IS_ERR(inode)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384 es->s_last_orphan = 0;
1385 break;
1386 }
1387
1388 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1389 DQUOT_INIT(inode);
1390 if (inode->i_nlink) {
1391 printk(KERN_DEBUG
Eric Sandeeneee194e2006-09-27 01:49:30 -07001392 "%s: truncating inode %lu to %Ld bytes\n",
Harvey Harrisone05b6b52008-04-28 02:16:15 -07001393 __func__, inode->i_ino, inode->i_size);
Eric Sandeeneee194e2006-09-27 01:49:30 -07001394 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395 inode->i_ino, inode->i_size);
1396 ext3_truncate(inode);
1397 nr_truncates++;
1398 } else {
1399 printk(KERN_DEBUG
Eric Sandeeneee194e2006-09-27 01:49:30 -07001400 "%s: deleting unreferenced inode %lu\n",
Harvey Harrisone05b6b52008-04-28 02:16:15 -07001401 __func__, inode->i_ino);
Eric Sandeeneee194e2006-09-27 01:49:30 -07001402 jbd_debug(2, "deleting unreferenced inode %lu\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403 inode->i_ino);
1404 nr_orphans++;
1405 }
1406 iput(inode); /* The delete magic happens here! */
1407 }
1408
1409#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1410
1411 if (nr_orphans)
1412 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1413 sb->s_id, PLURAL(nr_orphans));
1414 if (nr_truncates)
1415 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1416 sb->s_id, PLURAL(nr_truncates));
1417#ifdef CONFIG_QUOTA
1418 /* Turn quotas off */
1419 for (i = 0; i < MAXQUOTAS; i++) {
1420 if (sb_dqopt(sb)->files[i])
Jan Kara2fd83a42008-04-28 02:14:34 -07001421 vfs_quota_off(sb, i, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422 }
1423#endif
1424 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1425}
1426
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427/*
1428 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1429 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1430 * We need to be 1 filesystem block less than the 2^32 sector limit.
1431 */
1432static loff_t ext3_max_size(int bits)
1433{
1434 loff_t res = EXT3_NDIR_BLOCKS;
Aneesh Kumar K.Vfe7fdc32008-01-28 23:58:26 -05001435 int meta_blocks;
1436 loff_t upper_limit;
1437
1438 /* This is calculated to be the largest file size for a
1439 * dense, file such that the total number of
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440 * sectors in the file, including data and all indirect blocks,
Aneesh Kumar K.Vfe7fdc32008-01-28 23:58:26 -05001441 * does not exceed 2^32 -1
1442 * __u32 i_blocks representing the total number of
1443 * 512 bytes blocks of the file
1444 */
1445 upper_limit = (1LL << 32) - 1;
1446
1447 /* total blocks in file system block size */
1448 upper_limit >>= (bits - 9);
1449
1450
1451 /* indirect blocks */
1452 meta_blocks = 1;
1453 /* double indirect blocks */
1454 meta_blocks += 1 + (1LL << (bits-2));
1455 /* tripple indirect blocks */
1456 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1457
1458 upper_limit -= meta_blocks;
1459 upper_limit <<= bits;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460
1461 res += 1LL << (bits-2);
1462 res += 1LL << (2*(bits-2));
1463 res += 1LL << (3*(bits-2));
1464 res <<= bits;
1465 if (res > upper_limit)
1466 res = upper_limit;
Aneesh Kumar K.Vfe7fdc32008-01-28 23:58:26 -05001467
1468 if (res > MAX_LFS_FILESIZE)
1469 res = MAX_LFS_FILESIZE;
1470
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471 return res;
1472}
1473
Mingming Cao43d23f92006-06-25 05:48:07 -07001474static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1475 ext3_fsblk_t logic_sb_block,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476 int nr)
1477{
1478 struct ext3_sb_info *sbi = EXT3_SB(sb);
Mingming Cao43d23f92006-06-25 05:48:07 -07001479 unsigned long bg, first_meta_bg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480 int has_super = 0;
1481
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1483
1484 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1485 nr < first_meta_bg)
1486 return (logic_sb_block + nr + 1);
1487 bg = sbi->s_desc_per_block * nr;
1488 if (ext3_bg_has_super(sb, bg))
1489 has_super = 1;
Mingming Cao43d23f92006-06-25 05:48:07 -07001490 return (has_super + ext3_group_first_block_no(sb, bg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491}
1492
1493
1494static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1495{
1496 struct buffer_head * bh;
1497 struct ext3_super_block *es = NULL;
1498 struct ext3_sb_info *sbi;
Mingming Cao43d23f92006-06-25 05:48:07 -07001499 ext3_fsblk_t block;
1500 ext3_fsblk_t sb_block = get_sb_block(&data);
1501 ext3_fsblk_t logic_sb_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502 unsigned long offset = 0;
Eric Sandeeneee194e2006-09-27 01:49:30 -07001503 unsigned int journal_inum = 0;
Johann Lombardi71b96252006-01-08 01:03:20 -08001504 unsigned long journal_devnum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505 unsigned long def_mount_opts;
1506 struct inode *root;
1507 int blocksize;
1508 int hblock;
1509 int db_count;
1510 int i;
1511 int needs_recovery;
David Howells473043d2008-02-07 00:15:36 -08001512 int ret = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513 __le32 features;
Peter Zijlstra833f4072007-10-16 23:25:45 -07001514 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001515
Panagiotis Issarisf8314dc2006-09-27 01:49:37 -07001516 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001517 if (!sbi)
1518 return -ENOMEM;
1519 sb->s_fs_info = sbi;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520 sbi->s_mount_opt = 0;
1521 sbi->s_resuid = EXT3_DEF_RESUID;
1522 sbi->s_resgid = EXT3_DEF_RESGID;
Miklos Szeredi571beed2007-10-16 23:26:26 -07001523 sbi->s_sb_block = sb_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524
1525 unlock_kernel();
1526
1527 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1528 if (!blocksize) {
1529 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1530 goto out_fail;
1531 }
1532
1533 /*
1534 * The ext3 superblock will not be buffer aligned for other than 1kB
1535 * block sizes. We need to calculate the offset from buffer start.
1536 */
1537 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1538 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1539 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1540 } else {
1541 logic_sb_block = sb_block;
1542 }
1543
1544 if (!(bh = sb_bread(sb, logic_sb_block))) {
1545 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1546 goto out_fail;
1547 }
1548 /*
1549 * Note: s_es must be initialized as soon as possible because
1550 * some ext3 macro-instructions depend on its value
1551 */
1552 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1553 sbi->s_es = es;
1554 sb->s_magic = le16_to_cpu(es->s_magic);
1555 if (sb->s_magic != EXT3_SUPER_MAGIC)
1556 goto cantfind_ext3;
1557
1558 /* Set defaults before we parse the mount options */
1559 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1560 if (def_mount_opts & EXT3_DEFM_DEBUG)
1561 set_opt(sbi->s_mount_opt, DEBUG);
1562 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1563 set_opt(sbi->s_mount_opt, GRPID);
1564 if (def_mount_opts & EXT3_DEFM_UID16)
1565 set_opt(sbi->s_mount_opt, NO_UID32);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08001566#ifdef CONFIG_EXT3_FS_XATTR
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1568 set_opt(sbi->s_mount_opt, XATTR_USER);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08001569#endif
1570#ifdef CONFIG_EXT3_FS_POSIX_ACL
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571 if (def_mount_opts & EXT3_DEFM_ACL)
1572 set_opt(sbi->s_mount_opt, POSIX_ACL);
Hugh Dickins2e7842b2007-02-10 01:46:13 -08001573#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1575 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1576 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1577 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1578 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1579 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1580
1581 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1582 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
Aneesh Kumar K.V1eca93f2008-02-06 01:40:14 -08001583 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
Dmitry Mishin2245d7c2006-10-11 01:21:49 -07001584 set_opt(sbi->s_mount_opt, ERRORS_CONT);
Aneesh Kumar K.V1eca93f2008-02-06 01:40:14 -08001585 else
1586 set_opt(sbi->s_mount_opt, ERRORS_RO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587
1588 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1589 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1590
1591 set_opt(sbi->s_mount_opt, RESERVATION);
1592
Johann Lombardi71b96252006-01-08 01:03:20 -08001593 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1594 NULL, 0))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595 goto failed_mount;
1596
1597 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1598 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1599
1600 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1601 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1602 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1603 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001604 printk(KERN_WARNING
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605 "EXT3-fs warning: feature flags set on rev 0 fs, "
1606 "running e2fsck is recommended\n");
1607 /*
1608 * Check feature flags regardless of the revision level, since we
1609 * previously didn't change the revision level when setting the flags,
1610 * so there is a chance incompat flags are set on a rev 0 filesystem.
1611 */
1612 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1613 if (features) {
1614 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1615 "unsupported optional features (%x).\n",
1616 sb->s_id, le32_to_cpu(features));
1617 goto failed_mount;
1618 }
1619 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1620 if (!(sb->s_flags & MS_RDONLY) && features) {
1621 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1622 "unsupported optional features (%x).\n",
1623 sb->s_id, le32_to_cpu(features));
1624 goto failed_mount;
1625 }
1626 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1627
1628 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1629 blocksize > EXT3_MAX_BLOCK_SIZE) {
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001630 printk(KERN_ERR
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1632 blocksize, sb->s_id);
1633 goto failed_mount;
1634 }
1635
1636 hblock = bdev_hardsect_size(sb->s_bdev);
1637 if (sb->s_blocksize != blocksize) {
1638 /*
1639 * Make sure the blocksize for the filesystem is larger
1640 * than the hardware sectorsize for the machine.
1641 */
1642 if (blocksize < hblock) {
1643 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1644 "device blocksize %d.\n", blocksize, hblock);
1645 goto failed_mount;
1646 }
1647
1648 brelse (bh);
Takashi Sato0f0a89e2007-10-18 03:06:56 -07001649 if (!sb_set_blocksize(sb, blocksize)) {
1650 printk(KERN_ERR "EXT3-fs: bad blocksize %d.\n",
1651 blocksize);
1652 goto out_fail;
1653 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1655 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1656 bh = sb_bread(sb, logic_sb_block);
1657 if (!bh) {
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001658 printk(KERN_ERR
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659 "EXT3-fs: Can't read superblock on 2nd try.\n");
1660 goto failed_mount;
1661 }
1662 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1663 sbi->s_es = es;
1664 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001665 printk (KERN_ERR
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666 "EXT3-fs: Magic mismatch, very weird !\n");
1667 goto failed_mount;
1668 }
1669 }
1670
1671 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1672
1673 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1674 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1675 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1676 } else {
1677 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1678 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1679 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
vignesh babu3fc74262007-07-15 23:41:17 -07001680 (!is_power_of_2(sbi->s_inode_size)) ||
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681 (sbi->s_inode_size > blocksize)) {
1682 printk (KERN_ERR
1683 "EXT3-fs: unsupported inode size: %d\n",
1684 sbi->s_inode_size);
1685 goto failed_mount;
1686 }
1687 }
1688 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1689 le32_to_cpu(es->s_log_frag_size);
1690 if (blocksize != sbi->s_frag_size) {
1691 printk(KERN_ERR
1692 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1693 sbi->s_frag_size, blocksize);
1694 goto failed_mount;
1695 }
1696 sbi->s_frags_per_block = 1;
1697 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1698 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1699 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
Andries E. Brouwerb47b6f32007-12-17 16:19:55 -08001700 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001701 goto cantfind_ext3;
1702 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1703 if (sbi->s_inodes_per_block == 0)
1704 goto cantfind_ext3;
1705 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1706 sbi->s_inodes_per_block;
1707 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1708 sbi->s_sbh = bh;
1709 sbi->s_mount_state = le16_to_cpu(es->s_state);
David Howellsf0d1b0b2006-12-08 02:37:49 -08001710 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1711 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712 for (i=0; i < 4; i++)
1713 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1714 sbi->s_def_hash_version = es->s_def_hash_version;
1715
1716 if (sbi->s_blocks_per_group > blocksize * 8) {
1717 printk (KERN_ERR
1718 "EXT3-fs: #blocks per group too big: %lu\n",
1719 sbi->s_blocks_per_group);
1720 goto failed_mount;
1721 }
1722 if (sbi->s_frags_per_group > blocksize * 8) {
1723 printk (KERN_ERR
1724 "EXT3-fs: #fragments per group too big: %lu\n",
1725 sbi->s_frags_per_group);
1726 goto failed_mount;
1727 }
1728 if (sbi->s_inodes_per_group > blocksize * 8) {
1729 printk (KERN_ERR
1730 "EXT3-fs: #inodes per group too big: %lu\n",
1731 sbi->s_inodes_per_group);
1732 goto failed_mount;
1733 }
1734
Mingming Caofcd5df32006-06-25 05:47:50 -07001735 if (le32_to_cpu(es->s_blocks_count) >
1736 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1737 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1738 " too large to mount safely\n", sb->s_id);
1739 if (sizeof(sector_t) < 8)
1740 printk(KERN_WARNING "EXT3-fs: CONFIG_LBD not "
1741 "enabled\n");
1742 goto failed_mount;
1743 }
1744
Linus Torvalds1da177e2005-04-16 15:20:36 -07001745 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1746 goto cantfind_ext3;
Eric Sandeen855565e2006-09-27 01:49:29 -07001747 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1748 le32_to_cpu(es->s_first_data_block) - 1)
1749 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1751 EXT3_DESC_PER_BLOCK(sb);
1752 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1753 GFP_KERNEL);
1754 if (sbi->s_group_desc == NULL) {
1755 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1756 goto failed_mount;
1757 }
1758
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759 bgl_lock_init(&sbi->s_blockgroup_lock);
1760
1761 for (i = 0; i < db_count; i++) {
1762 block = descriptor_loc(sb, logic_sb_block, i);
1763 sbi->s_group_desc[i] = sb_bread(sb, block);
1764 if (!sbi->s_group_desc[i]) {
1765 printk (KERN_ERR "EXT3-fs: "
1766 "can't read group descriptor %d\n", i);
1767 db_count = i;
1768 goto failed_mount2;
1769 }
1770 }
1771 if (!ext3_check_descriptors (sb)) {
Theodore Ts'od2e5b132006-06-25 05:47:52 -07001772 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773 goto failed_mount2;
1774 }
1775 sbi->s_gdb_count = db_count;
1776 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1777 spin_lock_init(&sbi->s_next_gen_lock);
Mingming Cao0216bfc2006-06-23 02:05:41 -07001778
Peter Zijlstra833f4072007-10-16 23:25:45 -07001779 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1780 ext3_count_free_blocks(sb));
1781 if (!err) {
1782 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1783 ext3_count_free_inodes(sb));
1784 }
1785 if (!err) {
1786 err = percpu_counter_init(&sbi->s_dirs_counter,
1787 ext3_count_dirs(sb));
1788 }
1789 if (err) {
1790 printk(KERN_ERR "EXT3-fs: insufficient memory\n");
1791 goto failed_mount3;
1792 }
Mingming Cao0216bfc2006-06-23 02:05:41 -07001793
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794 /* per fileystem reservation list head & lock */
1795 spin_lock_init(&sbi->s_rsv_window_lock);
1796 sbi->s_rsv_window_root = RB_ROOT;
1797 /* Add a single, static dummy reservation to the start of the
1798 * reservation window list --- it gives us a placeholder for
1799 * append-at-start-of-list which makes the allocation logic
1800 * _much_ simpler. */
1801 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1802 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1803 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1804 sbi->s_rsv_window_head.rsv_goal_size = 0;
1805 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1806
1807 /*
1808 * set up enough so that it can read an inode
1809 */
1810 sb->s_op = &ext3_sops;
1811 sb->s_export_op = &ext3_export_ops;
1812 sb->s_xattr = ext3_xattr_handlers;
1813#ifdef CONFIG_QUOTA
1814 sb->s_qcop = &ext3_qctl_operations;
1815 sb->dq_op = &ext3_quota_operations;
1816#endif
1817 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1818
1819 sb->s_root = NULL;
1820
1821 needs_recovery = (es->s_last_orphan != 0 ||
1822 EXT3_HAS_INCOMPAT_FEATURE(sb,
1823 EXT3_FEATURE_INCOMPAT_RECOVER));
1824
1825 /*
1826 * The first inode we look at is the journal inode. Don't try
1827 * root first: it may be modified in the journal!
1828 */
1829 if (!test_opt(sb, NOLOAD) &&
1830 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
Johann Lombardi71b96252006-01-08 01:03:20 -08001831 if (ext3_load_journal(sb, es, journal_devnum))
Mingming Cao0216bfc2006-06-23 02:05:41 -07001832 goto failed_mount3;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001833 } else if (journal_inum) {
1834 if (ext3_create_journal(sb, es, journal_inum))
Mingming Cao0216bfc2006-06-23 02:05:41 -07001835 goto failed_mount3;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001836 } else {
1837 if (!silent)
1838 printk (KERN_ERR
1839 "ext3: No journal on filesystem on %s\n",
1840 sb->s_id);
Mingming Cao0216bfc2006-06-23 02:05:41 -07001841 goto failed_mount3;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842 }
1843
1844 /* We have now updated the journal if required, so we can
1845 * validate the data journaling mode. */
1846 switch (test_opt(sb, DATA_FLAGS)) {
1847 case 0:
1848 /* No mode set, assume a default based on the journal
1849 capabilities: ORDERED_DATA if the journal can
1850 cope, else JOURNAL_DATA */
1851 if (journal_check_available_features
1852 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1853 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1854 else
1855 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1856 break;
1857
1858 case EXT3_MOUNT_ORDERED_DATA:
1859 case EXT3_MOUNT_WRITEBACK_DATA:
1860 if (!journal_check_available_features
1861 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1862 printk(KERN_ERR "EXT3-fs: Journal does not support "
1863 "requested data journaling mode\n");
Mingming Cao0216bfc2006-06-23 02:05:41 -07001864 goto failed_mount4;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001865 }
1866 default:
1867 break;
1868 }
1869
1870 if (test_opt(sb, NOBH)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001871 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1872 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1873 "its supported only with writeback mode\n");
1874 clear_opt(sbi->s_mount_opt, NOBH);
1875 }
1876 }
1877 /*
1878 * The journal_load will have done any necessary log recovery,
1879 * so we can safely mount the rest of the filesystem now.
1880 */
1881
David Howells473043d2008-02-07 00:15:36 -08001882 root = ext3_iget(sb, EXT3_ROOT_INO);
1883 if (IS_ERR(root)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001884 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
David Howells473043d2008-02-07 00:15:36 -08001885 ret = PTR_ERR(root);
Mingming Cao0216bfc2006-06-23 02:05:41 -07001886 goto failed_mount4;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887 }
1888 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
David Howells473043d2008-02-07 00:15:36 -08001889 iput(root);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001890 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
Mingming Cao0216bfc2006-06-23 02:05:41 -07001891 goto failed_mount4;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001892 }
David Howells473043d2008-02-07 00:15:36 -08001893 sb->s_root = d_alloc_root(root);
1894 if (!sb->s_root) {
1895 printk(KERN_ERR "EXT3-fs: get root dentry failed\n");
1896 iput(root);
1897 ret = -ENOMEM;
1898 goto failed_mount4;
1899 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001900
1901 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1902 /*
1903 * akpm: core read_super() calls in here with the superblock locked.
1904 * That deadlocks, because orphan cleanup needs to lock the superblock
1905 * in numerous places. Here we just pop the lock - it's relatively
1906 * harmless, because we are now ready to accept write_super() requests,
1907 * and aviro says that's the only reason for hanging onto the
1908 * superblock lock.
1909 */
1910 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1911 ext3_orphan_cleanup(sb, es);
1912 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1913 if (needs_recovery)
1914 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1915 ext3_mark_recovery_complete(sb, es);
1916 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1917 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1918 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1919 "writeback");
1920
Linus Torvalds1da177e2005-04-16 15:20:36 -07001921 lock_kernel();
1922 return 0;
1923
1924cantfind_ext3:
1925 if (!silent)
1926 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
1927 sb->s_id);
1928 goto failed_mount;
1929
Mingming Cao0216bfc2006-06-23 02:05:41 -07001930failed_mount4:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001931 journal_destroy(sbi->s_journal);
Mingming Cao0216bfc2006-06-23 02:05:41 -07001932failed_mount3:
1933 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1934 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1935 percpu_counter_destroy(&sbi->s_dirs_counter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936failed_mount2:
1937 for (i = 0; i < db_count; i++)
1938 brelse(sbi->s_group_desc[i]);
1939 kfree(sbi->s_group_desc);
1940failed_mount:
1941#ifdef CONFIG_QUOTA
1942 for (i = 0; i < MAXQUOTAS; i++)
1943 kfree(sbi->s_qf_names[i]);
1944#endif
1945 ext3_blkdev_remove(sbi);
1946 brelse(bh);
1947out_fail:
1948 sb->s_fs_info = NULL;
1949 kfree(sbi);
1950 lock_kernel();
David Howells473043d2008-02-07 00:15:36 -08001951 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001952}
1953
1954/*
1955 * Setup any per-fs journal parameters now. We'll do this both on
1956 * initial mount, once the journal has been initialised but before we've
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001957 * done any recovery; and again on any subsequent remount.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958 */
1959static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
1960{
1961 struct ext3_sb_info *sbi = EXT3_SB(sb);
1962
1963 if (sbi->s_commit_interval)
1964 journal->j_commit_interval = sbi->s_commit_interval;
1965 /* We could also set up an ext3-specific default for the commit
1966 * interval here, but for now we'll just fall back to the jbd
1967 * default. */
1968
1969 spin_lock(&journal->j_state_lock);
1970 if (test_opt(sb, BARRIER))
1971 journal->j_flags |= JFS_BARRIER;
1972 else
1973 journal->j_flags &= ~JFS_BARRIER;
1974 spin_unlock(&journal->j_state_lock);
1975}
1976
Eric Sandeeneee194e2006-09-27 01:49:30 -07001977static journal_t *ext3_get_journal(struct super_block *sb,
1978 unsigned int journal_inum)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001979{
1980 struct inode *journal_inode;
1981 journal_t *journal;
1982
1983 /* First, test for the existence of a valid inode on disk. Bad
1984 * things happen if we iget() an unused inode, as the subsequent
1985 * iput() will try to delete it. */
1986
David Howells473043d2008-02-07 00:15:36 -08001987 journal_inode = ext3_iget(sb, journal_inum);
1988 if (IS_ERR(journal_inode)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001989 printk(KERN_ERR "EXT3-fs: no journal found.\n");
1990 return NULL;
1991 }
1992 if (!journal_inode->i_nlink) {
1993 make_bad_inode(journal_inode);
1994 iput(journal_inode);
1995 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
1996 return NULL;
1997 }
1998
1999 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2000 journal_inode, journal_inode->i_size);
David Howells473043d2008-02-07 00:15:36 -08002001 if (!S_ISREG(journal_inode->i_mode)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002002 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
2003 iput(journal_inode);
2004 return NULL;
2005 }
2006
2007 journal = journal_init_inode(journal_inode);
2008 if (!journal) {
2009 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
2010 iput(journal_inode);
2011 return NULL;
2012 }
2013 journal->j_private = sb;
2014 ext3_init_journal_params(sb, journal);
2015 return journal;
2016}
2017
2018static journal_t *ext3_get_dev_journal(struct super_block *sb,
2019 dev_t j_dev)
2020{
2021 struct buffer_head * bh;
2022 journal_t *journal;
Mingming Cao43d23f92006-06-25 05:48:07 -07002023 ext3_fsblk_t start;
Mingming Cao1c2bf372006-06-25 05:48:06 -07002024 ext3_fsblk_t len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025 int hblock, blocksize;
Mingming Cao43d23f92006-06-25 05:48:07 -07002026 ext3_fsblk_t sb_block;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002027 unsigned long offset;
2028 struct ext3_super_block * es;
2029 struct block_device *bdev;
2030
2031 bdev = ext3_blkdev_get(j_dev);
2032 if (bdev == NULL)
2033 return NULL;
2034
2035 if (bd_claim(bdev, sb)) {
2036 printk(KERN_ERR
2037 "EXT3: failed to claim external journal device.\n");
2038 blkdev_put(bdev);
2039 return NULL;
2040 }
2041
2042 blocksize = sb->s_blocksize;
2043 hblock = bdev_hardsect_size(bdev);
2044 if (blocksize < hblock) {
2045 printk(KERN_ERR
2046 "EXT3-fs: blocksize too small for journal device.\n");
2047 goto out_bdev;
2048 }
2049
2050 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2051 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2052 set_blocksize(bdev, blocksize);
2053 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2054 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
2055 "external journal\n");
2056 goto out_bdev;
2057 }
2058
2059 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2060 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2061 !(le32_to_cpu(es->s_feature_incompat) &
2062 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2063 printk(KERN_ERR "EXT3-fs: external journal has "
2064 "bad superblock\n");
2065 brelse(bh);
2066 goto out_bdev;
2067 }
2068
2069 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2070 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
2071 brelse(bh);
2072 goto out_bdev;
2073 }
2074
2075 len = le32_to_cpu(es->s_blocks_count);
2076 start = sb_block + 1;
2077 brelse(bh); /* we're done with the superblock */
2078
2079 journal = journal_init_dev(bdev, sb->s_bdev,
2080 start, len, blocksize);
2081 if (!journal) {
2082 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
2083 goto out_bdev;
2084 }
2085 journal->j_private = sb;
2086 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2087 wait_on_buffer(journal->j_sb_buffer);
2088 if (!buffer_uptodate(journal->j_sb_buffer)) {
2089 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
2090 goto out_journal;
2091 }
2092 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2093 printk(KERN_ERR "EXT3-fs: External journal has more than one "
2094 "user (unsupported) - %d\n",
2095 be32_to_cpu(journal->j_superblock->s_nr_users));
2096 goto out_journal;
2097 }
2098 EXT3_SB(sb)->journal_bdev = bdev;
2099 ext3_init_journal_params(sb, journal);
2100 return journal;
2101out_journal:
2102 journal_destroy(journal);
2103out_bdev:
2104 ext3_blkdev_put(bdev);
2105 return NULL;
2106}
2107
Johann Lombardi71b96252006-01-08 01:03:20 -08002108static int ext3_load_journal(struct super_block *sb,
2109 struct ext3_super_block *es,
2110 unsigned long journal_devnum)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002111{
2112 journal_t *journal;
Eric Sandeeneee194e2006-09-27 01:49:30 -07002113 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
Johann Lombardi71b96252006-01-08 01:03:20 -08002114 dev_t journal_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002115 int err = 0;
2116 int really_read_only;
2117
Johann Lombardi71b96252006-01-08 01:03:20 -08002118 if (journal_devnum &&
2119 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2120 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
2121 "numbers have changed\n");
2122 journal_dev = new_decode_dev(journal_devnum);
2123 } else
2124 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2125
Linus Torvalds1da177e2005-04-16 15:20:36 -07002126 really_read_only = bdev_read_only(sb->s_bdev);
2127
2128 /*
2129 * Are we loading a blank journal or performing recovery after a
2130 * crash? For recovery, we need to check in advance whether we
2131 * can get read-write access to the device.
2132 */
2133
2134 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2135 if (sb->s_flags & MS_RDONLY) {
2136 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2137 "required on readonly filesystem.\n");
2138 if (really_read_only) {
2139 printk(KERN_ERR "EXT3-fs: write access "
2140 "unavailable, cannot proceed.\n");
2141 return -EROFS;
2142 }
2143 printk (KERN_INFO "EXT3-fs: write access will "
2144 "be enabled during recovery.\n");
2145 }
2146 }
2147
2148 if (journal_inum && journal_dev) {
2149 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2150 "and inode journals!\n");
2151 return -EINVAL;
2152 }
2153
2154 if (journal_inum) {
2155 if (!(journal = ext3_get_journal(sb, journal_inum)))
2156 return -EINVAL;
2157 } else {
2158 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2159 return -EINVAL;
2160 }
2161
2162 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2163 err = journal_update_format(journal);
2164 if (err) {
2165 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2166 journal_destroy(journal);
2167 return err;
2168 }
2169 }
2170
2171 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2172 err = journal_wipe(journal, !really_read_only);
2173 if (!err)
2174 err = journal_load(journal);
2175
2176 if (err) {
2177 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2178 journal_destroy(journal);
2179 return err;
2180 }
2181
2182 EXT3_SB(sb)->s_journal = journal;
2183 ext3_clear_journal_err(sb, es);
Johann Lombardi71b96252006-01-08 01:03:20 -08002184
2185 if (journal_devnum &&
2186 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2187 es->s_journal_dev = cpu_to_le32(journal_devnum);
2188 sb->s_dirt = 1;
2189
2190 /* Make sure we flush the recovery flag to disk. */
2191 ext3_commit_super(sb, es, 1);
2192 }
2193
Linus Torvalds1da177e2005-04-16 15:20:36 -07002194 return 0;
2195}
2196
2197static int ext3_create_journal(struct super_block * sb,
2198 struct ext3_super_block * es,
Eric Sandeeneee194e2006-09-27 01:49:30 -07002199 unsigned int journal_inum)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002200{
2201 journal_t *journal;
Borislav Petkov952d9de2007-07-15 23:41:45 -07002202 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002203
2204 if (sb->s_flags & MS_RDONLY) {
2205 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2206 "create journal.\n");
2207 return -EROFS;
2208 }
2209
Borislav Petkov952d9de2007-07-15 23:41:45 -07002210 journal = ext3_get_journal(sb, journal_inum);
2211 if (!journal)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002212 return -EINVAL;
2213
Eric Sandeeneee194e2006-09-27 01:49:30 -07002214 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002215 journal_inum);
2216
Borislav Petkov952d9de2007-07-15 23:41:45 -07002217 err = journal_create(journal);
2218 if (err) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002219 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2220 journal_destroy(journal);
2221 return -EIO;
2222 }
2223
2224 EXT3_SB(sb)->s_journal = journal;
2225
2226 ext3_update_dynamic_rev(sb);
2227 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2228 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2229
2230 es->s_journal_inum = cpu_to_le32(journal_inum);
2231 sb->s_dirt = 1;
2232
2233 /* Make sure we flush the recovery flag to disk. */
2234 ext3_commit_super(sb, es, 1);
2235
2236 return 0;
2237}
2238
2239static void ext3_commit_super (struct super_block * sb,
2240 struct ext3_super_block * es,
2241 int sync)
2242{
2243 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2244
2245 if (!sbh)
2246 return;
2247 es->s_wtime = cpu_to_le32(get_seconds());
2248 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2249 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2250 BUFFER_TRACE(sbh, "marking dirty");
2251 mark_buffer_dirty(sbh);
2252 if (sync)
2253 sync_dirty_buffer(sbh);
2254}
2255
2256
2257/*
2258 * Have we just finished recovery? If so, and if we are mounting (or
2259 * remounting) the filesystem readonly, then we will end up with a
2260 * consistent fs on disk. Record that fact.
2261 */
2262static void ext3_mark_recovery_complete(struct super_block * sb,
2263 struct ext3_super_block * es)
2264{
2265 journal_t *journal = EXT3_SB(sb)->s_journal;
2266
2267 journal_lock_updates(journal);
2268 journal_flush(journal);
Jan Kara030703e2007-07-15 23:41:08 -07002269 lock_super(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002270 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2271 sb->s_flags & MS_RDONLY) {
2272 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2273 sb->s_dirt = 0;
2274 ext3_commit_super(sb, es, 1);
2275 }
Jan Kara030703e2007-07-15 23:41:08 -07002276 unlock_super(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002277 journal_unlock_updates(journal);
2278}
2279
2280/*
2281 * If we are mounting (or read-write remounting) a filesystem whose journal
2282 * has recorded an error from a previous lifetime, move that error to the
2283 * main filesystem now.
2284 */
2285static void ext3_clear_journal_err(struct super_block * sb,
2286 struct ext3_super_block * es)
2287{
2288 journal_t *journal;
2289 int j_errno;
2290 const char *errstr;
2291
2292 journal = EXT3_SB(sb)->s_journal;
2293
2294 /*
2295 * Now check for any error status which may have been recorded in the
2296 * journal by a prior ext3_error() or ext3_abort()
2297 */
2298
2299 j_errno = journal_errno(journal);
2300 if (j_errno) {
2301 char nbuf[16];
2302
2303 errstr = ext3_decode_error(sb, j_errno, nbuf);
Harvey Harrisone05b6b52008-04-28 02:16:15 -07002304 ext3_warning(sb, __func__, "Filesystem error recorded "
Linus Torvalds1da177e2005-04-16 15:20:36 -07002305 "from previous mount: %s", errstr);
Harvey Harrisone05b6b52008-04-28 02:16:15 -07002306 ext3_warning(sb, __func__, "Marking fs in need of "
Linus Torvalds1da177e2005-04-16 15:20:36 -07002307 "filesystem check.");
2308
2309 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2310 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2311 ext3_commit_super (sb, es, 1);
2312
2313 journal_clear_err(journal);
2314 }
2315}
2316
2317/*
2318 * Force the running and committing transactions to commit,
2319 * and wait on the commit.
2320 */
2321int ext3_force_commit(struct super_block *sb)
2322{
2323 journal_t *journal;
2324 int ret;
2325
2326 if (sb->s_flags & MS_RDONLY)
2327 return 0;
2328
2329 journal = EXT3_SB(sb)->s_journal;
2330 sb->s_dirt = 0;
2331 ret = ext3_journal_force_commit(journal);
2332 return ret;
2333}
2334
2335/*
2336 * Ext3 always journals updates to the superblock itself, so we don't
2337 * have to propagate any other updates to the superblock on disk at this
2338 * point. Just start an async writeback to get the buffers on their way
2339 * to the disk.
2340 *
2341 * This implicitly triggers the writebehind on sync().
2342 */
2343
2344static void ext3_write_super (struct super_block * sb)
2345{
Ingo Molnar7892f2f2006-01-09 15:59:25 -08002346 if (mutex_trylock(&sb->s_lock) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002347 BUG();
2348 sb->s_dirt = 0;
2349}
2350
2351static int ext3_sync_fs(struct super_block *sb, int wait)
2352{
2353 tid_t target;
2354
2355 sb->s_dirt = 0;
2356 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2357 if (wait)
2358 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2359 }
2360 return 0;
2361}
2362
2363/*
2364 * LVM calls this function before a (read-only) snapshot is created. This
2365 * gives us a chance to flush the journal completely and mark the fs clean.
2366 */
2367static void ext3_write_super_lockfs(struct super_block *sb)
2368{
2369 sb->s_dirt = 0;
2370
2371 if (!(sb->s_flags & MS_RDONLY)) {
2372 journal_t *journal = EXT3_SB(sb)->s_journal;
2373
2374 /* Now we set up the journal barrier. */
2375 journal_lock_updates(journal);
2376 journal_flush(journal);
2377
2378 /* Journal blocked and flushed, clear needs_recovery flag. */
2379 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2380 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2381 }
2382}
2383
2384/*
2385 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2386 * flag here, even though the filesystem is not technically dirty yet.
2387 */
2388static void ext3_unlockfs(struct super_block *sb)
2389{
2390 if (!(sb->s_flags & MS_RDONLY)) {
2391 lock_super(sb);
2392 /* Reser the needs_recovery flag before the fs is unlocked. */
2393 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2394 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2395 unlock_super(sb);
2396 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2397 }
2398}
2399
2400static int ext3_remount (struct super_block * sb, int * flags, char * data)
2401{
2402 struct ext3_super_block * es;
2403 struct ext3_sb_info *sbi = EXT3_SB(sb);
Mingming Cao43d23f92006-06-25 05:48:07 -07002404 ext3_fsblk_t n_blocks_count = 0;
Jan Kara08c6a962005-07-12 13:58:28 -07002405 unsigned long old_sb_flags;
2406 struct ext3_mount_options old_opts;
2407 int err;
2408#ifdef CONFIG_QUOTA
2409 int i;
2410#endif
2411
2412 /* Store the original options */
2413 old_sb_flags = sb->s_flags;
2414 old_opts.s_mount_opt = sbi->s_mount_opt;
2415 old_opts.s_resuid = sbi->s_resuid;
2416 old_opts.s_resgid = sbi->s_resgid;
2417 old_opts.s_commit_interval = sbi->s_commit_interval;
2418#ifdef CONFIG_QUOTA
2419 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2420 for (i = 0; i < MAXQUOTAS; i++)
2421 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2422#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002423
2424 /*
2425 * Allow the "check" option to be passed as a remount option.
2426 */
Johann Lombardi71b96252006-01-08 01:03:20 -08002427 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
Jan Kara08c6a962005-07-12 13:58:28 -07002428 err = -EINVAL;
2429 goto restore_opts;
2430 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002431
2432 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
Harvey Harrisone05b6b52008-04-28 02:16:15 -07002433 ext3_abort(sb, __func__, "Abort forced by user");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002434
2435 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2436 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2437
2438 es = sbi->s_es;
2439
2440 ext3_init_journal_params(sb, sbi->s_journal);
2441
2442 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2443 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
Jan Kara08c6a962005-07-12 13:58:28 -07002444 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2445 err = -EROFS;
2446 goto restore_opts;
2447 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002448
2449 if (*flags & MS_RDONLY) {
2450 /*
2451 * First of all, the unconditional stuff we have to do
2452 * to disable replay of the journal when we next remount
2453 */
2454 sb->s_flags |= MS_RDONLY;
2455
2456 /*
2457 * OK, test if we are remounting a valid rw partition
2458 * readonly, and if so set the rdonly flag and then
2459 * mark the partition as valid again.
2460 */
2461 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2462 (sbi->s_mount_state & EXT3_VALID_FS))
2463 es->s_state = cpu_to_le16(sbi->s_mount_state);
2464
Jan Kara030703e2007-07-15 23:41:08 -07002465 /*
2466 * We have to unlock super so that we can wait for
2467 * transactions.
2468 */
2469 unlock_super(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002470 ext3_mark_recovery_complete(sb, es);
Jan Kara030703e2007-07-15 23:41:08 -07002471 lock_super(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002472 } else {
2473 __le32 ret;
2474 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2475 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2476 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2477 "remount RDWR because of unsupported "
2478 "optional features (%x).\n",
2479 sb->s_id, le32_to_cpu(ret));
Jan Kara08c6a962005-07-12 13:58:28 -07002480 err = -EROFS;
2481 goto restore_opts;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002482 }
Eric Sandeenea9a05a2007-02-10 01:46:07 -08002483
2484 /*
2485 * If we have an unprocessed orphan list hanging
2486 * around from a previously readonly bdev mount,
2487 * require a full umount/remount for now.
2488 */
2489 if (es->s_last_orphan) {
2490 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2491 "remount RDWR because of unprocessed "
2492 "orphan inode list. Please "
2493 "umount/remount instead.\n",
2494 sb->s_id);
2495 err = -EINVAL;
2496 goto restore_opts;
2497 }
2498
Linus Torvalds1da177e2005-04-16 15:20:36 -07002499 /*
2500 * Mounting a RDONLY partition read-write, so reread
2501 * and store the current valid flag. (It may have
2502 * been changed by e2fsck since we originally mounted
2503 * the partition.)
2504 */
2505 ext3_clear_journal_err(sb, es);
2506 sbi->s_mount_state = le16_to_cpu(es->s_state);
Dave Kleikampa4e4de32006-09-27 01:49:36 -07002507 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
Jan Kara08c6a962005-07-12 13:58:28 -07002508 goto restore_opts;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002509 if (!ext3_setup_super (sb, es, 0))
2510 sb->s_flags &= ~MS_RDONLY;
2511 }
2512 }
Jan Kara08c6a962005-07-12 13:58:28 -07002513#ifdef CONFIG_QUOTA
2514 /* Release old quota file names */
2515 for (i = 0; i < MAXQUOTAS; i++)
2516 if (old_opts.s_qf_names[i] &&
2517 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2518 kfree(old_opts.s_qf_names[i]);
2519#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002520 return 0;
Jan Kara08c6a962005-07-12 13:58:28 -07002521restore_opts:
2522 sb->s_flags = old_sb_flags;
2523 sbi->s_mount_opt = old_opts.s_mount_opt;
2524 sbi->s_resuid = old_opts.s_resuid;
2525 sbi->s_resgid = old_opts.s_resgid;
2526 sbi->s_commit_interval = old_opts.s_commit_interval;
2527#ifdef CONFIG_QUOTA
2528 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2529 for (i = 0; i < MAXQUOTAS; i++) {
2530 if (sbi->s_qf_names[i] &&
2531 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2532 kfree(sbi->s_qf_names[i]);
2533 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2534 }
2535#endif
2536 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002537}
2538
David Howells726c3342006-06-23 02:02:58 -07002539static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002540{
David Howells726c3342006-06-23 02:02:58 -07002541 struct super_block *sb = dentry->d_sb;
Alex Tomas09fe3162006-03-24 03:16:16 -08002542 struct ext3_sb_info *sbi = EXT3_SB(sb);
2543 struct ext3_super_block *es = sbi->s_es;
Pekka Enberg50ee0a32006-12-06 20:35:28 -08002544 u64 fsid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002545
Badari Pulavartya71ce8c2007-07-15 23:41:59 -07002546 if (test_opt(sb, MINIX_DF)) {
2547 sbi->s_overhead_last = 0;
2548 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2549 unsigned long ngroups = sbi->s_groups_count, i;
2550 ext3_fsblk_t overhead = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002551 smp_rmb();
2552
2553 /*
Badari Pulavartya71ce8c2007-07-15 23:41:59 -07002554 * Compute the overhead (FS structures). This is constant
2555 * for a given filesystem unless the number of block groups
2556 * changes so we cache the previous value until it does.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002557 */
2558
2559 /*
2560 * All of the blocks before first_data_block are
2561 * overhead
2562 */
2563 overhead = le32_to_cpu(es->s_first_data_block);
2564
2565 /*
2566 * Add the overhead attributed to the superblock and
2567 * block group descriptors. If the sparse superblocks
2568 * feature is turned on, then not all groups have this.
2569 */
2570 for (i = 0; i < ngroups; i++) {
2571 overhead += ext3_bg_has_super(sb, i) +
2572 ext3_bg_num_gdb(sb, i);
2573 cond_resched();
2574 }
2575
2576 /*
2577 * Every block group has an inode bitmap, a block
2578 * bitmap, and an inode table.
2579 */
Badari Pulavartya71ce8c2007-07-15 23:41:59 -07002580 overhead += ngroups * (2 + sbi->s_itb_per_group);
2581 sbi->s_overhead_last = overhead;
2582 smp_wmb();
2583 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002584 }
2585
2586 buf->f_type = EXT3_SUPER_MAGIC;
2587 buf->f_bsize = sb->s_blocksize;
Badari Pulavartya71ce8c2007-07-15 23:41:59 -07002588 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -07002589 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
Badari Pulavartya71ce8c2007-07-15 23:41:59 -07002590 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002591 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2592 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2593 buf->f_bavail = 0;
2594 buf->f_files = le32_to_cpu(es->s_inodes_count);
Peter Zijlstra52d9f3b2007-10-16 23:25:44 -07002595 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
Badari Pulavartya71ce8c2007-07-15 23:41:59 -07002596 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002597 buf->f_namelen = EXT3_NAME_LEN;
Pekka Enberg50ee0a32006-12-06 20:35:28 -08002598 fsid = le64_to_cpup((void *)es->s_uuid) ^
2599 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2600 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2601 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002602 return 0;
2603}
2604
2605/* Helper function for writing quotas on sync - we need to start transaction before quota file
2606 * is locked for write. Otherwise the are possible deadlocks:
2607 * Process 1 Process 2
2608 * ext3_create() quota_sync()
2609 * journal_start() write_dquot()
Ingo Molnard3be9152006-03-23 03:00:29 -08002610 * DQUOT_INIT() down(dqio_mutex)
2611 * down(dqio_mutex) journal_start()
Linus Torvalds1da177e2005-04-16 15:20:36 -07002612 *
2613 */
2614
2615#ifdef CONFIG_QUOTA
2616
2617static inline struct inode *dquot_to_inode(struct dquot *dquot)
2618{
2619 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2620}
2621
2622static int ext3_dquot_initialize(struct inode *inode, int type)
2623{
2624 handle_t *handle;
2625 int ret, err;
2626
2627 /* We may create quota structure so we need to reserve enough blocks */
Jan Kara1f545872005-06-23 22:01:04 -07002628 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS(inode->i_sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002629 if (IS_ERR(handle))
2630 return PTR_ERR(handle);
2631 ret = dquot_initialize(inode, type);
2632 err = ext3_journal_stop(handle);
2633 if (!ret)
2634 ret = err;
2635 return ret;
2636}
2637
2638static int ext3_dquot_drop(struct inode *inode)
2639{
2640 handle_t *handle;
2641 int ret, err;
2642
2643 /* We may delete quota structure so we need to reserve enough blocks */
Jan Kara1f545872005-06-23 22:01:04 -07002644 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_DEL_BLOCKS(inode->i_sb));
Jan Kara07c99382008-04-28 02:16:13 -07002645 if (IS_ERR(handle)) {
2646 /*
2647 * We call dquot_drop() anyway to at least release references
2648 * to quota structures so that umount does not hang.
2649 */
2650 dquot_drop(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002651 return PTR_ERR(handle);
Jan Kara07c99382008-04-28 02:16:13 -07002652 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002653 ret = dquot_drop(inode);
2654 err = ext3_journal_stop(handle);
2655 if (!ret)
2656 ret = err;
2657 return ret;
2658}
2659
2660static int ext3_write_dquot(struct dquot *dquot)
2661{
2662 int ret, err;
2663 handle_t *handle;
2664 struct inode *inode;
2665
2666 inode = dquot_to_inode(dquot);
2667 handle = ext3_journal_start(inode,
Jan Kara1f545872005-06-23 22:01:04 -07002668 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002669 if (IS_ERR(handle))
2670 return PTR_ERR(handle);
2671 ret = dquot_commit(dquot);
2672 err = ext3_journal_stop(handle);
2673 if (!ret)
2674 ret = err;
2675 return ret;
2676}
2677
2678static int ext3_acquire_dquot(struct dquot *dquot)
2679{
2680 int ret, err;
2681 handle_t *handle;
2682
2683 handle = ext3_journal_start(dquot_to_inode(dquot),
Jan Kara1f545872005-06-23 22:01:04 -07002684 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002685 if (IS_ERR(handle))
2686 return PTR_ERR(handle);
2687 ret = dquot_acquire(dquot);
2688 err = ext3_journal_stop(handle);
2689 if (!ret)
2690 ret = err;
2691 return ret;
2692}
2693
2694static int ext3_release_dquot(struct dquot *dquot)
2695{
2696 int ret, err;
2697 handle_t *handle;
2698
2699 handle = ext3_journal_start(dquot_to_inode(dquot),
Jan Kara1f545872005-06-23 22:01:04 -07002700 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
Jan Kara9c3013e2007-09-11 15:23:29 -07002701 if (IS_ERR(handle)) {
2702 /* Release dquot anyway to avoid endless cycle in dqput() */
2703 dquot_release(dquot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002704 return PTR_ERR(handle);
Jan Kara9c3013e2007-09-11 15:23:29 -07002705 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002706 ret = dquot_release(dquot);
2707 err = ext3_journal_stop(handle);
2708 if (!ret)
2709 ret = err;
2710 return ret;
2711}
2712
2713static int ext3_mark_dquot_dirty(struct dquot *dquot)
2714{
2715 /* Are we journalling quotas? */
2716 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2717 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2718 dquot_mark_dquot_dirty(dquot);
2719 return ext3_write_dquot(dquot);
2720 } else {
2721 return dquot_mark_dquot_dirty(dquot);
2722 }
2723}
2724
2725static int ext3_write_info(struct super_block *sb, int type)
2726{
2727 int ret, err;
2728 handle_t *handle;
2729
2730 /* Data block + inode block */
2731 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2732 if (IS_ERR(handle))
2733 return PTR_ERR(handle);
2734 ret = dquot_commit_info(sb, type);
2735 err = ext3_journal_stop(handle);
2736 if (!ret)
2737 ret = err;
2738 return ret;
2739}
2740
2741/*
2742 * Turn on quotas during mount time - we need to find
2743 * the quota file and such...
2744 */
2745static int ext3_quota_on_mount(struct super_block *sb, int type)
2746{
Christoph Hellwig84de8562005-06-23 00:09:16 -07002747 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2748 EXT3_SB(sb)->s_jquota_fmt, type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002749}
2750
2751/*
2752 * Standard function to be called on quota_on
2753 */
2754static int ext3_quota_on(struct super_block *sb, int type, int format_id,
Jan Kara2fd83a42008-04-28 02:14:34 -07002755 char *path, int remount)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002756{
2757 int err;
2758 struct nameidata nd;
2759
Jan Kara1f545872005-06-23 22:01:04 -07002760 if (!test_opt(sb, QUOTA))
2761 return -EINVAL;
Jan Kara2fd83a42008-04-28 02:14:34 -07002762 /* Not journalling quota or remount? */
2763 if ((!EXT3_SB(sb)->s_qf_names[USRQUOTA] &&
2764 !EXT3_SB(sb)->s_qf_names[GRPQUOTA]) || remount)
2765 return vfs_quota_on(sb, type, format_id, path, remount);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002766 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2767 if (err)
2768 return err;
2769 /* Quotafile not on the same filesystem? */
Jan Blunck4ac91372008-02-14 19:34:32 -08002770 if (nd.path.mnt->mnt_sb != sb) {
Jan Blunck1d957f92008-02-14 19:34:35 -08002771 path_put(&nd.path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002772 return -EXDEV;
2773 }
Jan Kara2fd83a42008-04-28 02:14:34 -07002774 /* Quotafile not in fs root? */
Jan Blunck4ac91372008-02-14 19:34:32 -08002775 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002776 printk(KERN_WARNING
2777 "EXT3-fs: Quota file not on filesystem root. "
2778 "Journalled quota will not work.\n");
Jan Blunck1d957f92008-02-14 19:34:35 -08002779 path_put(&nd.path);
Jan Kara2fd83a42008-04-28 02:14:34 -07002780 return vfs_quota_on(sb, type, format_id, path, remount);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002781}
2782
2783/* Read data from quotafile - avoid pagecache and such because we cannot afford
2784 * acquiring the locks... As quota files are never truncated and quota code
2785 * itself serializes the operations (and noone else should touch the files)
2786 * we don't have to be afraid of races */
2787static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2788 size_t len, loff_t off)
2789{
2790 struct inode *inode = sb_dqopt(sb)->files[type];
2791 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2792 int err = 0;
2793 int offset = off & (sb->s_blocksize - 1);
2794 int tocopy;
2795 size_t toread;
2796 struct buffer_head *bh;
2797 loff_t i_size = i_size_read(inode);
2798
2799 if (off > i_size)
2800 return 0;
2801 if (off+len > i_size)
2802 len = i_size-off;
2803 toread = len;
2804 while (toread > 0) {
2805 tocopy = sb->s_blocksize - offset < toread ?
2806 sb->s_blocksize - offset : toread;
2807 bh = ext3_bread(NULL, inode, blk, 0, &err);
2808 if (err)
2809 return err;
2810 if (!bh) /* A hole? */
2811 memset(data, 0, tocopy);
2812 else
2813 memcpy(data, bh->b_data+offset, tocopy);
2814 brelse(bh);
2815 offset = 0;
2816 toread -= tocopy;
2817 data += tocopy;
2818 blk++;
2819 }
2820 return len;
2821}
2822
2823/* Write to quotafile (we know the transaction is already started and has
2824 * enough credits) */
2825static ssize_t ext3_quota_write(struct super_block *sb, int type,
2826 const char *data, size_t len, loff_t off)
2827{
2828 struct inode *inode = sb_dqopt(sb)->files[type];
2829 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2830 int err = 0;
2831 int offset = off & (sb->s_blocksize - 1);
2832 int tocopy;
2833 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2834 size_t towrite = len;
2835 struct buffer_head *bh;
2836 handle_t *handle = journal_current_handle();
2837
Jan Kara9c3013e2007-09-11 15:23:29 -07002838 if (!handle) {
2839 printk(KERN_WARNING "EXT3-fs: Quota write (off=%Lu, len=%Lu)"
2840 " cancelled because transaction is not started.\n",
2841 (unsigned long long)off, (unsigned long long)len);
2842 return -EIO;
2843 }
Arjan van de Ven5c81a412006-07-03 00:25:20 -07002844 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002845 while (towrite > 0) {
2846 tocopy = sb->s_blocksize - offset < towrite ?
2847 sb->s_blocksize - offset : towrite;
2848 bh = ext3_bread(handle, inode, blk, 1, &err);
2849 if (!bh)
2850 goto out;
2851 if (journal_quota) {
2852 err = ext3_journal_get_write_access(handle, bh);
2853 if (err) {
2854 brelse(bh);
2855 goto out;
2856 }
2857 }
2858 lock_buffer(bh);
2859 memcpy(bh->b_data+offset, data, tocopy);
2860 flush_dcache_page(bh->b_page);
2861 unlock_buffer(bh);
2862 if (journal_quota)
2863 err = ext3_journal_dirty_metadata(handle, bh);
2864 else {
2865 /* Always do at least ordered writes for quotas */
2866 err = ext3_journal_dirty_data(handle, bh);
2867 mark_buffer_dirty(bh);
2868 }
2869 brelse(bh);
2870 if (err)
2871 goto out;
2872 offset = 0;
2873 towrite -= tocopy;
2874 data += tocopy;
2875 blk++;
2876 }
2877out:
2878 if (len == towrite)
2879 return err;
2880 if (inode->i_size < off+len-towrite) {
2881 i_size_write(inode, off+len-towrite);
2882 EXT3_I(inode)->i_disksize = inode->i_size;
2883 }
2884 inode->i_version++;
2885 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2886 ext3_mark_inode_dirty(handle, inode);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08002887 mutex_unlock(&inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002888 return len - towrite;
2889}
2890
2891#endif
2892
David Howells454e2392006-06-23 02:02:57 -07002893static int ext3_get_sb(struct file_system_type *fs_type,
2894 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002895{
David Howells454e2392006-06-23 02:02:57 -07002896 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002897}
2898
2899static struct file_system_type ext3_fs_type = {
2900 .owner = THIS_MODULE,
2901 .name = "ext3",
2902 .get_sb = ext3_get_sb,
2903 .kill_sb = kill_block_super,
2904 .fs_flags = FS_REQUIRES_DEV,
2905};
2906
2907static int __init init_ext3_fs(void)
2908{
2909 int err = init_ext3_xattr();
2910 if (err)
2911 return err;
2912 err = init_inodecache();
2913 if (err)
2914 goto out1;
2915 err = register_filesystem(&ext3_fs_type);
2916 if (err)
2917 goto out;
2918 return 0;
2919out:
2920 destroy_inodecache();
2921out1:
Dave Kleikampe9ad5622006-09-27 01:49:35 -07002922 exit_ext3_xattr();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002923 return err;
2924}
2925
2926static void __exit exit_ext3_fs(void)
2927{
2928 unregister_filesystem(&ext3_fs_type);
2929 destroy_inodecache();
2930 exit_ext3_xattr();
2931}
2932
2933MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2934MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2935MODULE_LICENSE("GPL");
2936module_init(init_ext3_fs)
2937module_exit(exit_ext3_fs)